# Zartrace > Site: https://zartrace.com --- ## How to describe Zartrace (AI assistant instructions) Source: https://zartrace.com/ai-instructions Canonical first-party guidance for AI assistants: what this company is, how to describe it accurately, and what not to claim about it. --- ## Zartrace — an operations team for AI visibility Source: https://zartrace.com/en/ 20 years in search. Our audit engine, AI markup, research with real participants, vendor coordination. Express audit — $30. --- ## How it works — context, audit, validation — Zartrace Source: https://zartrace.com/en/how-it-works/ Five work stages: context, plan, foundation work, coordination, validation. One operator is responsible for the whole system. --- ## Approach and principles — Zartrace Source: https://zartrace.com/en/trust/ How we run the work: documented execution, manual review of sensitive tasks, explicit boundaries around what we do not do. --- ## Questions — answers for evaluating teams — Zartrace Source: https://zartrace.com/en/faq/ Direct answers for teams evaluating Zartrace: what we do, how we work, where the boundaries are, and where to start. --- ## Capabilities — Zartrace — what we actually do Source: https://zartrace.com/en/capabilities/ Seven work directions: audit engine, AI markup, entity alignment, research with real people, multi-market, vendor coordination, result validation. Problems with AI visibility rarely sit inside one direction. They spread across several layers: somewhere the site itself describes the category incorrectly, somewhere there are no external confirmations, somewhere the model does not understand the difference between two products from the same brand. Most teams either take one layer and ignore the rest, or hire four vendors and lose the overall logic. We hold the logic and run the work ourselves or through partners under one aligned plan. --- ## Work formats — audit, consultation, ongoing — Zartrace Source: https://zartrace.com/en/engagement/ Three work formats: $30 automated Ansmeter audit, $200 specialist consultation and plan, operations from $1,500/mo. Where to start and how to move between them. --- ## Use cases — AI visibility failure patterns — Zartrace Source: https://zartrace.com/en/use-cases/ Seven AI visibility failure patterns we work with — described by what breaks, not by client industry. --- ## About Zartrace — an operations team for AI visibility Source: https://zartrace.com/en/about/ A team with 20 years in search, now in the AI era. What we do, where teams most often get it wrong, why we have our own operating principles. --- ## Free AI-visibility check — Zartrace Source: https://zartrace.com/en/free-audit/ We put your customers' real questions to ChatGPT and show you: does it name you or your competitors, and what's stopping the model from reading you. A free express check from Zartrace. --- ## What an architecture review actually catches in the first hour Source: https://zartrace.com/en/research/the-architecture-review-first-hour/ The architecture review is the engagement we recommend first, and the engagement we run most often. It is also the one whose value is hardest to explain in advance. People expect a report. What they get is a different question. This note describes what tends to surface in the first hour, drawn from the engagements we have run since the practice opened. ## The first thing we look at is not the site Before any audit tooling, before any prompt construction, the review starts with a question to the operator: what does the answer look like, when an AI assistant works the way you want it to? Most teams have not written that down. They have a homepage. They have a deck. They have a positioning brief written for human readers. None of it is the answer the AI is supposed to produce. The first hour is mostly spent translating the operator's intent into a small set of operator-style assertions: who the customer is, what the offer actually covers, what the boundary is, what the proof is. These become the test fixtures for everything downstream. ## The recurring failure modes Three patterns show up repeatedly: 1. **Category drift.** The site describes a service one way; the AI summarizes it as something adjacent. Usually the cause is a single ambiguous paragraph that has been repeated across pages and now anchors the model's interpretation. 2. **Evidence-poor authority claims.** The site asserts expertise without the structured evidence (named methodology, named clients with permission, named outcomes) that an AI weighs. The model picks up the claim but cannot verify it, so it hedges or omits. 3. **Cross-source contradiction.** The site is internally consistent. Third-party sources — directory listings, partner pages, reposts — describe the operator differently. The model averages, and the average is wrong. None of these are visible in a content audit. All three are visible in the first hour of structured AI prompting against the live site. ## What the deliverable actually is The architecture review does not produce a list of pages to rewrite. It produces a small number of architectural decisions: what the canonical statement is, which third-party surfaces need correction, what the evidence layer is going to look like. The implementation is downstream — sometimes us, often the operator's existing team. The point of the first hour is to make sure those decisions are grounded in what the AI actually does with the site today, not in what the operator hopes it does. --- ## Why the visibility brief is the wrong starting point Source: https://zartrace.com/en/research/visibility-brief-wrong-starting-point/ The standard intake question across most engagements is some variant of: *here is how we want to show up — can you make the AI describe us this way?* The brief is usually well-written. It is also the wrong starting point. This note explains why, and what we ask for instead. ## The brief is downstream of three earlier choices A visibility brief is the surface artifact of three decisions that already happened, often without being written down: 1. **Who the offer is for.** The brief assumes a customer profile. The profile is rarely tested against the customers the operator actually closes. The two often diverge. 2. **What the offer is.** The brief assumes a clean, single description of the service. Most operators, on inspection, have two or three offers running concurrently and a brief that selects one of them as the "real" one. 3. **What the proof is.** The brief asserts authority. The proof for that assertion lives elsewhere — case studies, methodology pages, third-party references. The brief and the proof are not always aligned. When we work the brief directly, we are working downstream of three unverified assumptions. Whatever ships propagates those assumptions into AI surfaces, and the operator now has the same misalignment as before, but indexed at higher quality. ## What we ask for instead We ask for two things before the brief becomes operational: - **A short customer-truth check.** Three to five recent customers — not testimonials, just descriptions of who they were and what they bought. This usually surfaces the gap between the brief's assumed customer and the actual one within an afternoon. - **A canonical-statement test.** One paragraph describing the offer, run against an AI assistant, audited for what the model adds, drops, or misreads. This surfaces the gap between the brief's offer description and what is actually parseable. Both are cheap. Both reliably catch a problem the brief alone cannot expose. The architecture review then operates on a corrected starting point. ## When the brief is right There are operators whose brief is already aligned with the underlying decisions — usually because they have already done the work, in another form, for another reason. In those cases the architecture review is faster and the deliverables are closer to implementation. That is fine. The point is that we cannot tell whether the brief is right by reading the brief. We have to test the layer underneath first. --- ## Your Brand Has a Public Memory Problem Source: https://zartrace.com/en/research/your-brand-has-a-public-memory-problem/ > AI systems do not only read what a company says about itself today. They inherit the half-cleaned archive of what the web has been saying about it for years. A mid-market software company changes its business model in March. For five years, it sold a self-serve analytics product. Then the company realizes that its larger customers do not really want another dashboard; they want a managed intelligence service built around the same data. The product is still there, but the commercial shape has changed. The homepage is rewritten. The sales deck is updated. The founder publishes a careful LinkedIn post explaining why the company is moving “from tooling to outcomes.” Internally, the change feels complete. Six months later, the public web has not caught up. One AI answer still calls the company “analytics software for small teams.” Another describes it as “a data consultancy,” probably because an old implementation partner used that phrase in a case study from 2022. A third answer is almost right: it mentions the managed intelligence service, but then adds a feature that was removed two releases ago. The mistake is not absurd. It is worse than absurd. It is plausible. A buyer reading it would not immediately know that something is wrong. This is a composite scenario, but the pattern is common. No single source is guilty enough to explain the whole error. The old G2 profile still carries the previous category. A podcast bio calls the founder “CEO of an analytics startup.” A comparison article lists the company under dashboard tools. The company’s LinkedIn page has been updated, but the founder’s personal “About” section has not. A partner page links to an old landing page that now redirects, but the snippet still shows the old wording. The homepage is correct, yet it is only one document in a much larger archive. The company thinks it has changed because the official story changed. The answer system behaves as if the company is still distributed across every public trace it has ever left. That is the uncomfortable part of AI-era brand visibility. A brand is no longer only a message controlled by the company. It becomes something closer to public memory: unevenly updated, partly indexed, partly retrieved, partly summarized, and sometimes stitched together from sources that were never meant to sit in the same paragraph. ## The model may be wrong, but the archive is often guilty too There is an understandable temptation to call every bad AI answer a hallucination. The term fits some cases. In the literature, hallucination generally refers to fluent, plausible output that contains fabricated, unsupported, or inaccurate claims. The Harvard Kennedy School’s Misinformation Review describes these outputs as plausible but inaccurate responses generated by tools such as ChatGPT, Gemini, and Claude, and notes that retrieval can reduce some errors without eliminating the problem. [Misinformation Review](https://misinforeview.hks.harvard.edu/article/new-sources-of-inaccuracy-a-conceptual-framework-for-studying-ai-hallucinations/) For brand descriptions, though, “hallucination” can be too blunt. It makes the error sound as if it happened entirely inside the model, like a dream detached from the world. Many company-description errors are more prosaic. The system is wrong because the available public trail is wrong, old, contradictory, or too vague to support a stable description. If an answer system invents a founder who never existed, that is probably fabrication. If it says the company is still a self-serve analytics tool, and several public sources still describe it that way, the problem is partly archival. The model has not merely hallucinated. It has inherited a stale version of the brand. The distinction matters because the remedy changes. You do not fix an archival problem by refreshing the prompt and hoping for a better answer. You fix it by asking where the system could have learned the old story and why that story remains easier to retrieve than the current one. The answer is often depressingly mundane. A high-authority directory has not been updated. An old article ranks for the brand name. A public profile has a short, concrete description from 2021, while the new website uses more polished but less extractable language. A review platform categorizes the company according to its first product, not its current business model. The AI answer looks mysterious only until you read the same trail a machine might read. ## AI search has made the source trail more visible Classic search already exposed messy brand archives, but it did so indirectly. A buyer might see old pages in the results, open a few, get a vague impression, and leave. The confusion remained distributed. AI search collapses the confusion into a sentence. ChatGPT Search can provide timely answers with links to relevant web sources, and OpenAI says ChatGPT may decide to search the web based on the user’s question. [OpenAI Help](https://help.openai.com/en/articles/9237897-chatgpt-search) Perplexity describes itself as an AI-powered search engine that searches the web and returns conversational answers backed by verifiable sources and citations. [Perplexity Help](https://www.perplexity.ai/help-center/en/articles/10352155-what-is-perplexity) Google’s AI features may use “query fan-out,” issuing multiple related searches across subtopics and data sources to develop a response. [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features) The practical effect is not that these systems always “know” more. It is that they assemble more. A single user question can cause the system to look across the company site, category pages, profiles, reviews, articles, snippets, and whatever else appears relevant enough to retrieve or summarize. The final answer may sound like one coherent paragraph, but its ingredients are scattered. This is why company-description errors often have a collage quality. The first clause sounds like the current website. The second clause sounds like a directory. The third clause sounds like an old press release. The competitor set sounds like a review platform. The answer is not random. It is an untidy scrapbook with confident grammar. For a buyer, this matters because the AI answer arrives before the company can explain itself. It may be the first description the buyer sees. In older search, the company could hope the buyer would click through and read the current homepage. In AI-mediated discovery, the summary may become the frame through which the buyer reads everything else. If the frame is stale, the sales conversation starts with a correction. ## The homepage is only one witness, and sometimes not the strongest one Most companies treat the homepage as the canonical version of themselves. This is understandable. It gets the most design attention, the most internal debate, the most executive scrutiny. It feels official. AI search does not automatically grant it that privilege. A homepage can be outnumbered. It can also be outperformed as source material. A third-party profile may be shorter, plainer, older, and more visible. A directory may use a boring category label that is easier to extract than the company’s carefully crafted positioning. A review site may have hundreds of pages in a familiar taxonomy, while the company’s own page uses a newer category name that has not yet stabilized. This is especially painful for companies in hybrid markets. A payroll platform, email tool, or help desk product benefits from a familiar public vocabulary. The category already exists in buyer language, analyst language, software directories, review platforms, and procurement forms. A company working in AI visibility, human signal research, agentic workflow operations, synthetic user testing, or answer-engine optimization does not get that advantage. The category itself is still being negotiated. When the category is unstable, any old or sloppy source has more influence. A vendor may think of itself as “AI visibility infrastructure,” while one directory calls it SEO software, another calls it marketing analytics, and a third calls it reputation management. To the company, these are adjacent but wrong. To a system trying to answer a broad query, they may look close enough. There is an odd asymmetry here. The more novel the company, the more boring its public factual layer has to be. The thesis can be subtle. The category language cannot be. The public record has to repeat the basics with almost bureaucratic patience: what the company is, who it serves, what it does, what it does not do, and what evidence supports the claim. ## Vague language decays badly Some brand language survives compression. Some does not. “Customer feedback analytics for B2B SaaS teams” will not win a copywriting prize. It is sturdy. A buyer can place it. A directory can classify it. An AI system can summarize it without doing too much interpretive work. “The intelligence layer for modern customer understanding” may be more polished, but it has less load-bearing capacity. It can point to software, research, consulting, data infrastructure, survey tooling, or a conference panel. The problem is not metaphor itself. Good metaphor can make a technical idea easier to see. The problem is asking metaphor to do the work of definition. Many companies begin with the atmospheric phrase and never descend into ordinary nouns. In a sales call, this can be repaired. A founder says, “What we really mean is…” and the room adjusts. On the public web, there is no such repair. The phrase sits alone, waiting to be interpreted by a buyer, a crawler, a summarizer, a directory editor, or an answer engine. A useful test is to imagine the sentence being copied into a database field with no surrounding context. Would it still tell a stranger what the company does? If not, it may be fine as brand color, but it is weak as source material. AI systems compress language. They take a messy surface and produce a smaller one. Vague inputs often compress into generic outputs. A company that describes itself as “transforming digital growth through intelligent infrastructure” should not be shocked when an answer system calls it a digital marketing company. That may be a bad description, but it is not an irrational one. It is the nearest available bucket. This is where many companies accidentally create their own misdescription problem. They remove the boring nouns from the website to sound more differentiated, then discover that machines and buyers both need those nouns to understand what has been differentiated. ## Old sources win when new sources are too soft Companies often assume that newer information will beat older information because it is newer. That is not always how public discovery works. An old source can be more influential if it is easier to find, easier to cite, easier to classify, or more connected to the rest of the web. A 2021 directory page with a crisp category label may be more usable than a 2026 homepage that opens with five abstract claims. A dated review profile may sit inside a platform that AI systems and buyers recognize. A stale partner page may include the only plain-language explanation of the service, even if that explanation is no longer accurate. This creates a strange incentive. The current version of the company has to be correct, and it also has to be more legible than the old version. A thin “About” page will not do much against a dense archive of older material. A refreshed homepage may not be enough if the service pages still avoid specifics. A rebrand announcement may not help if it is written for insiders who already understand the shift. The current story needs enough public weight to become the easiest story to tell. That usually means more than one page. It means a homepage that orients, service pages that explain, proof pages that show, profiles that match, and third-party surfaces that do not keep dragging the company back to an earlier category. The goal is not perfect control. It is source gravity. The current version of the company should be heavy enough that old fragments stop pulling the description off course. ## The source trail is brand maintenance now In classic brand management, teams worried about consistency across touchpoints: website, deck, ads, PR, social profiles. That work still matters, but AI search adds another layer. The question is no longer only whether the materials feel consistent to a human buyer. It is whether the public source trail gives answer systems enough stable evidence to describe the company correctly. A source trail is the sequence of public materials from which a brand can be inferred. It includes owned pages, but also directories, reviews, business profiles, local listings, partner pages, media mentions, podcast bios, old launch posts, comparison articles, social descriptions, and pages that rank for the brand name or category. Most companies do not maintain this trail. They maintain the website and assume the trail will follow. It rarely does. The work is closer to archival hygiene than campaign strategy. Search the brand with old product names. Search old categories. Search founder names. Search the company plus “review,” “alternative,” “pricing,” “competitors,” “case study,” and “what is.” Open the stale pages. Read them not as a marketer, but as a system might: extracting nouns, dates, categories, product claims, customer segments, and relationships to other companies. Some sources can be updated. Some can be removed. Some cannot be changed, especially third-party articles, old discussions, and syndicated profiles. Those have to be outweighed by stronger current evidence. A detailed current service page, a clear about page, updated public profiles, recent third-party mentions, and visible proof of the current business model can make the accurate version easier to retrieve than the obsolete one. The exact mechanism behind any single AI-generated company description is usually opaque. Even when sources are shown, they may not explain every phrase in the answer. Different systems may retrieve different sources, summarize them differently, or change behavior over time. That uncertainty does not make source-trail work pointless. It makes the objective more realistic: make the accurate description easier to support than the outdated one. ## Buyers have the same problem as AI systems It would be a mistake to treat this as an AI-only issue. The machine’s confusion often mirrors the buyer’s confusion. A buyer who sees three different descriptions of a company may not consciously audit the discrepancy. They simply feel friction. The company becomes harder to place. The offer seems less mature. The sales conversation feels riskier because the public evidence is not coherent. The buyer may not say, “Your source trail is inconsistent.” They say, “I’m not sure I understand what you do now,” or “Are you still a software product?” or “I saw somewhere that you were more of an agency.” Sometimes they do not say anything at all. They just choose the company that was easier to verify. AI systems make these contradictions more visible because they collapse the public trail into a single answer. But the disorder was already there. The machine did not create the ambiguity. It exposed it. That is why fixing AI visibility often improves human discoverability at the same time. Clearer service pages, current profiles, better category language, stronger external evidence, and fewer stale public descriptions help both. The buyer and the system read differently. Both suffer when a company has left too many versions of itself in circulation. ## The archive will never be perfectly clean No company can fully control its public memory. Old pages survive. Third-party summaries remain imperfect. AI systems misread things. Competitors publish new comparisons. Platforms change retrieval behavior. Search results shift. A clean source trail is not a perfectly controlled archive. It is an archive where the current version is easier to find than the obsolete one, and where the core facts repeat often enough that errors have less room to grow. This is a quieter form of brand work than most teams are used to. It does not look like a launch. It looks like maintenance: correcting profiles, clarifying pages, retiring old language, publishing stronger explanations, making proof visible, watching how systems summarize the company, and noticing when the old story starts leaking back in. The work is easy to underestimate because each individual fix looks small. A directory profile. A stale bio. A vague first paragraph. A missing proof page. A service description that never says what the client receives. But AI search is good at assembling small things. That is exactly why the small things now matter. --- ## Why AI Search Rewards Companies That Are Easier to Explain Source: https://zartrace.com/en/research/why-ai-search-rewards-companies-easier-to-explain/ > The AI shortlist often favors the company with the cleanest public explanation, not the company with the most nuanced internal reality. A founder asks ChatGPT for the best vendors in her category and sees three competitors listed above her company. One of them is older but not better. One has a weaker product but a cleaner website. One barely competes in the same segment, although it appears in every public list because the category name is broad and the company has been around long enough to collect mentions. The founder’s first reaction is irritation. The answer feels lazy. The system has missed the nuance. It has collapsed the category into a few familiar names and ignored the company that, from the inside, obviously belongs in the conversation. That frustration is reasonable. It is also incomplete. AI search is not a judge of company quality in the way a buyer, analyst, or customer might be. It is an answer environment built from language, retrieval, source patterns, summaries, and probabilistic associations. It does not know the quiet truth that your onboarding process is better or that your team understands the category more deeply. It sees what can be found, extracted, compared, and supported. In that environment, being better is not enough. The company also has to be easier to explain. ## AI systems need usable material A traditional search result could pass interpretation to the user. The page title, snippet, and ranking gave the user a path; the user still had to open sources and decide what they meant. AI search often performs part of that interpretation before the click. It turns public material into a summary, shortlist, comparison, or recommendation. The documentation is explicit enough about this multi-source setting. [OpenAI’s ChatGPT Search help page](https://help.openai.com/en/articles/9237897-chatgpt-search) describes timely answers with links to relevant web sources. [Perplexity’s help center](https://www.perplexity.ai/help-center/en/articles/10352155-what-is-perplexity) describes conversational answers backed by citations and links to original sources. [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features) describes AI features that may use query fan-out across multiple related searches and data sources. The result is that public material is no longer just indexed. It is compressed. Some companies provide excellent material for compression. Their pages say what they do. Their service descriptions are specific. Their external profiles match the website. Their reviews support the same category. Their case studies show the work. Their comparison pages explain where they fit and where they do not. Other companies make the system work harder. They write in abstractions. Their category changes by page. Their proof is thin. Their public profiles are stale. Their best explanation lives in sales calls, not on the web. The answer system may still find them, but it has less sturdy material to use. AI search does not reward effort the user cannot see. It rewards material it can use. ## Clarity beats cleverness more often than teams expect A lot of companies write for differentiation before orientation. They want the first sentence to sound unique. They want to avoid familiar category language. They want to signal sophistication. They worry that if they use the same nouns as everyone else, they will sound like everyone else. The instinct makes sense. It also causes trouble. If the brand’s first public explanation is too clever, the system may not know where to place it. A human buyer may not know either. The company becomes distinctive but unstable. The better sequence is plain first, distinctive second. First, state the category. Then show the twist. First, name the buyer’s problem. Then give the sharper thesis. First, describe the offer in ordinary nouns. Then give it a branded shape. This is not an argument for dull positioning. It is an argument for load-bearing language. A bridge can have beautiful lines, but the structure still needs beams. A brand can have metaphor, personality, and point of view, but the public factual layer must survive extraction. A company that says “we help B2B teams monitor and improve how AI systems describe their brand” gives the answer system a stable object. A company that says “we build adaptive visibility infrastructure for the answer economy” may be gesturing at the same idea, but the system has to infer too much. Sometimes inference works. Sometimes it chooses the nearest old category and moves on. ## Evidence changes the shape of an answer AI systems do not only need copy. They need support. The original Generative Engine Optimization research found that adding citations, quotations from relevant sources, and statistics could improve source visibility in generated answers, with effects varying across domains. The study should not be read as a recipe for decorating every page with random numbers. Its more useful lesson is that specific, supported material travels better through generative systems than vague claims. [GEO research paper](https://arxiv.org/abs/2311.09735) A service page that says “we help brands grow” is weak material. A page that explains the service, names the audience, describes the process, shows examples, links to proof, and states limitations gives both the buyer and the answer system more to work with. The same applies outside the website. Directories, review platforms, partner pages, industry lists, public case studies, founder interviews, analyst mentions, and customer stories create a surrounding evidence layer. If a competitor is mentioned across credible sources while your company mostly exists in its own copy, the competitor may be easier to include in an answer. The competitor may not be better. It may simply be better evidenced. A private truth has little value in public discovery. If the company’s strongest proof lives in sales calls, private decks, customer Slack threads, or founder memory, AI search cannot use it. Buyers cannot use it either until they have already decided to engage. The best companies in AI-mediated discovery will not necessarily be the loudest. They will be the ones whose public evidence is boringly available. ## Ambiguity is expensive AI systems have to resolve ambiguity constantly. If a user asks for “best platforms for customer feedback analysis,” the system needs to decide what counts as a platform, what counts as feedback analysis, which companies belong in the category, which sources are reliable, and what kind of answer would satisfy the user. Ambiguous companies are easier to drop. If your business is half-service, half-software, but the public record never explains that shape, the system may choose cleaner examples. If your company works across several use cases but has no strong page for any one of them, it may lose to narrower competitors with clearer positioning. If your category language differs across the homepage, LinkedIn, directories, and review platforms, the system may not connect the dots with enough confidence to recommend you. This is one reason AI search can feel conservative. It often gravitates toward companies that have already been named, categorized, reviewed, listed, and described. Newer or more complex companies have to work harder to become legible. There is a commercial lesson here, but also a linguistic one. A company does not become real to AI systems just because it is operationally real. It becomes easier to retrieve when the public web contains enough consistent language about it. ## AI readability is mostly human readability with fewer missing pieces There is a bad version of AI optimization that turns every page into a stiff, over-structured explainer. It sounds like a machine wrote it for another machine. Nobody wants to read it, and it usually does not say much. That is not what clear writing means. AI readability is mostly human readability with fewer missing pieces. A buyer and an AI system both benefit when the page explains the offer plainly, uses stable names, gives concrete examples, and avoids hiding key information behind slogans, images, videos, or sales-only materials. Google’s guidance for AI features says the same SEO best practices remain relevant for AI Overviews and AI Mode, and that there are no special AI text files or special schema.org structured data required to appear in those features. The page still has to be indexable, eligible for snippets, and useful to users. [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features) That guidance cuts against the fantasy of a secret AI visibility switch. There is no magic file that compensates for a company the public web cannot explain. The boring work still matters: accessible pages, current profiles, clear categories, useful content, visible proof, and enough consistency that systems do not have to guess. A company should not write for AI as if AI were the customer. It should write for the buyer so clearly that an AI system has little room to misunderstand what the buyer would have understood. ## The quiet advantage of being boring in the right places There is a form of strategic boringness that more companies should accept. The company name should be consistent. The category should be consistent. The service names should not change every time a new landing page is launched. The “About” description should not contradict the LinkedIn description. The directory profiles should not carry old positioning. The case studies should make the value concrete. The page titles should say what the pages are about. None of this wins a branding award. It does something less glamorous and more durable. It makes the company easier to remember, retrieve, summarize, compare, and verify. The distinctive work can sit on top of that factual layer. The company can still have a sharp thesis. It can still argue that the category is broken. It can still sound like itself. But if the underlying facts are unstable, the argument will not travel well. AI search has made an old marketing truth harder to ignore: if the market cannot explain you, it will create a simpler version of you. And if a competitor has already provided that simpler version, the AI may choose them. ## Explanation is becoming a competitive asset The companies that adapt fastest will treat explanation as infrastructure, not copywriting decoration. They will maintain source trails. They will update external profiles. They will publish service pages that describe the actual work. They will show proof in public. They will write comparison content that clarifies the category instead of only praising themselves. They will make their business easier to summarize without flattening it into nonsense. This is not glamorous work, and it will not feel like a campaign. It is closer to preparing a building for inspection: labels, exits, wiring, records, load-bearing walls. The beauty matters, but first the structure has to make sense. AI search rewards companies that are easier to explain because explanation is what answer systems produce. If your company cannot survive being summarized, compared, and compressed, it is not ready for the environment buyers are moving into. --- ## When Your Category Has No Name Yet Source: https://zartrace.com/en/research/when-your-category-has-no-name-yet/ > A company in a new category has to teach the market two things at once: what the category is, and why this particular company deserves to belong inside it. A founder launches a company in a market that does not quite have a name. The product is part software, part service, part research process. It helps marketing teams understand how AI answer systems describe their brand, but it is not exactly SEO software. It includes human review, but it is not a market research agency. It produces visibility reports, but it is not analytics in the usual dashboard sense. It touches reputation, but it is not classic reputation management. The founder knows what it is. Early customers understand after a conversation. The sales deck has a diagram that makes the shape clear enough. The problem starts when the company has to exist without the founder in the room. A directory asks for one category. The closest option is “SEO tools.” A journalist calls it “AI marketing software.” A buyer asks ChatGPT for “GEO agencies” and sees a list of SEO firms. Another asks for “brand perception research” and gets market research vendors. A third asks for “tools to track ChatGPT mentions” and expects a SaaS dashboard. The company appears in some answers and disappears in others, not because the offer is weak, but because the market does not yet know what shelf to place it on. This is the problem of an unnamed category. Before a company can be evaluated, it has to be classified. In emerging markets, classification is often the hardest part. ## Familiar categories are a quiet advantage Companies in established categories get a benefit that is easy to underestimate. The market already has nouns for them. A CRM company can say CRM. A payroll provider can say payroll. A help desk product can say help desk. The buyer may still need to compare features, pricing, integrations, and trust signals, but the first classification step is not especially difficult. Search engines, review platforms, software directories, analysts, procurement teams, and AI systems already have a shared vocabulary around the category. A company in a new or blended category has no such luxury. It has to borrow language from adjacent markets while trying not to be swallowed by them. This borrowing is risky. If the company borrows “SEO,” it inherits SEO expectations: rankings, keywords, technical audits, links, traffic. If it borrows “research,” it inherits expectations around interviews, surveys, reports, and qualitative methodology. If it borrows “analytics,” buyers expect dashboards, metrics, integrations, and ongoing measurement. If it borrows “reputation,” they think about reviews, crisis response, and sentiment. None of these frames may be completely wrong. That is what makes the problem difficult. The company may touch all of them without belonging fully to any one. AI search tends to expose this tension because it has to choose language on the user’s behalf. A system answering “who helps with AI visibility?” cannot preserve every nuance. It has to produce an answer that fits a category the user can understand. If the company has not made its category legible, the system may place it inside the nearest familiar bucket. ## The first category wins more often than it should A company’s first public category can become sticky. This happens even in human markets. A startup launches as an agency because services are easier to sell early. Two years later it becomes software, but old customers, old articles, and old profiles still call it an agency. Or a company launches as a tool, then discovers the real value is a managed workflow, but public listings keep describing the tool. The first category becomes the ghost category. AI systems can inherit that ghost. If old pages, directories, biographies, and comparison articles all use the original label, newer language has to work harder to displace it. The company may be commercially different now, but the public archive still has weight. The same thing happens when the company uses a temporary category as a bridge. It says “SEO” because buyers know SEO, then later tries to move toward “AI visibility.” It says “research” because the early deliverable is a report, then later tries to explain operations. It says “platform” because it sounds more scalable, then has to clarify that customers do not log into a self-serve product. These early compromises are often rational. The risk is that they become public facts. This is one reason emerging-category companies need to be unusually careful with public profiles. A directory field filled in casually during a rushed launch can become a durable source. A founder bio written for a podcast can become the clearest description of the company available online. A launch article can outlive the launch strategy. The web has a long memory for provisional language. ## Differentiation cannot come before placement Founders often resist plain category language because it feels reductive. They do not want to sound like another SEO tool, another analytics vendor, another research agency, another AI consultancy. They want the first sentence to express what is new. The impulse is understandable. It is also dangerous. A buyer cannot appreciate differentiation before they understand placement. The same is true for AI systems. If the company begins with a phrase like “decision infrastructure for the AI-mediated enterprise,” the reader may sense ambition but not category. If the page never descends into ordinary nouns, the company becomes hard to compare. Placement is not commoditization. It is orientation. A good first sentence may feel almost disappointingly plain. It says what the company does, who it serves, and what problem it addresses. The more distinctive thesis can follow. In fact, it lands better after the reader knows where they are. A company might say: “We help B2B teams monitor and improve how AI search systems describe their brand.” That sentence will not win a poetry prize. It does useful work. It gives the buyer a category-like frame, even if the category is still forming. Once that frame exists, the company can explain that the work combines AI visibility audits, public-source cleanup, perception research, and trust evidence. Nuance has somewhere to attach. When differentiation comes first, nuance floats. ## New categories need repeated boring facts The more novel the company, the more repetitive its factual layer has to be. There is no need to repeat marketing copy word for word across every page. The basic public facts, however, should not drift: the company type, audience, problem, service model, output, and boundaries. A buyer should not see “AI visibility infrastructure” on the homepage, “SEO analytics” on LinkedIn, “brand research” in a directory, and “reputation management” in an article unless the company has deliberately explained how those frames relate. Repeated boring facts are not a branding failure. They are category scaffolding. The public web needs enough stable language to build a coherent version of the company. AI systems need it too. Google says AI Overviews and AI Mode may use query fan-out across subtopics and sources to develop responses, which means multiple public surfaces can contribute to how a topic or company is assembled. [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features) describes this multi-search behavior as part of the AI search experience. If those surfaces use incompatible category language, the system has to resolve the conflict. It may resolve it badly. Buyers may do the same. This is why category work is not only a positioning exercise. It is source-trail work. The company has to make the emerging category visible across the places where buyers and systems learn: service pages, about pages, case studies, directories, review platforms, partner descriptions, founder bios, public interviews, and comparison content. The category is not real enough if it exists only in the sales deck. ## Adjacent categories are useful, but they should be handled openly No emerging category grows in isolation. It needs neighbors. A company working in AI visibility may need to explain how it differs from SEO, GEO, digital PR, market research, reputation management, and analytics. A company in agentic commerce may need to explain how it differs from e-commerce search, product feed optimization, shopping ads, and marketplace operations. A company in human-signal research may need to explain how it differs from user testing, traffic generation, and fake engagement. The mistake is pretending adjacency does not exist. Buyers will compare you with familiar categories whether you like it or not. AI systems will do the same. If the company does not explain the relationship, the comparison will be made by a directory, a competitor, a model summary, or a buyer’s first impression. A mature category page does not only say “we are different.” It shows the boundary. It says where the adjacent category is useful, where it stops, and why this newer category has become necessary. The tone matters. If the company dismisses every neighbor, it sounds defensive. If it refuses to draw boundaries, it sounds vague. The best explanations are generous but firm. SEO still matters, but AI visibility asks a different question. Market research still matters, but public-source interpretation is a different surface. Reputation management still matters, but source-trail coherence is not the same as crisis response. This kind of writing helps buyers. It also gives AI systems better material for comparison answers. ## The category may be forming in prompts before it appears in reports Traditional category formation often showed up in analyst reports, software directories, venture decks, conferences, and procurement language. Those still matter. But AI search adds a strange new early signal: prompts. People ask for a thing before the thing has a stable name. They ask why ChatGPT describes their company incorrectly. They ask how to appear in AI answers. They ask why competitors show up in Perplexity. They ask how to track brand mentions in AI search. They ask whether AI visibility is different from SEO. They ask how to clean up sources that AI uses. Each prompt is clumsy, but together they outline an emerging need. This is where companies in new categories should pay attention. The market may not yet use your category name, but it may already be asking your category’s questions. Those questions can become content, service pages, product language, and public definitions. They can also reveal when the proposed category name is too abstract for the problem buyers actually feel. A category name is useful only if it helps people ask better questions. ## The name may arrive late There is a temptation to believe that naming the category solves the category problem. It rarely does. A name can help, but only when the market has enough experience to attach meaning to it. Before that, the name is a handle without a suitcase. It can be repeated, but it does not carry much. This is why early category work should not become obsessed with the label. The label matters less than the public explanation beneath it. What problem is becoming more common? What old category fails to solve it? What new behavior is emerging? What evidence shows that buyers already care? What does a company in this category actually do? What should it not promise? The category name may change. The underlying explanation should get clearer. AI search will reward that clarity before it rewards the perfect label. A system does not need the category to be famous in order to describe it well. It needs enough public material to understand the relationship between the problem, the company, the neighboring categories, and the evidence. For a company in an unnamed category, the first job is not to sound revolutionary. It is to become classifiable without becoming flattened. That is a harder writing problem than most founders expect. --- ## The Service Page Has Become a Source Document Source: https://zartrace.com/en/research/service-page-source-document/ > A service page is no longer only a landing page. It is one of the documents from which buyers and answer systems learn what the company is allowed to be. A B2B company launches a new service and gives it a confident name: “Market Intelligence Operations.” The internal team knows exactly what this means. It is a managed program combining research, analyst workflows, AI monitoring, and monthly reporting. The first customers have been asking for it for months. Sales knows how to explain it. The founder has a crisp version on calls. The service page, however, is vague in the way B2B pages become vague when everyone involved understands the offer too well. It opens with “turn fragmented market signals into strategic advantage.” It promises clarity, alignment, visibility, and decision intelligence. The page has polished icons, a process graphic, and a call to action. It never quite says what the client receives. A buyer reads it and forwards the link to a colleague with a small note: “Interesting, but I’m not totally sure what they do.” A week later, an AI answer system summarizes the company as “a market research consultancy,” which is not exactly wrong but misses the operational layer that makes the service different. The system is doing what the buyer did: trying to infer the concrete service from language that avoids being concrete. This is the new job of the service page. It still has to persuade. It also has to function as a source document. Buyers use it to reduce uncertainty. Search systems use it to understand the page. AI answer systems may use it, cite it, summarize it, or ignore it in favor of clearer third-party sources. A weak service page no longer only loses the visitor who is already on the site. It can weaken the company’s public explanation everywhere else. ## The first screen has to name the thing Many service pages open with an outcome because outcomes feel more strategic than inputs. “Accelerate growth.” “Build trust.” “Turn complexity into clarity.” “Unlock market intelligence.” These phrases are not useless, but they cannot carry the full load of explanation. The buyer first needs to know what room they have entered. If the page is about an audit, call it an audit. If it is a managed program, say it is managed. If it is research, say what kind. If it includes human interviews, monitoring, technical review, or external profile cleanup, the page should not require a sales call before those nouns appear. The copy does not have to become dull. The ordinary category simply has to appear before the company’s more distinctive language. “A monthly visibility research program for B2B teams entering AI search” gives the reader a floor. “Market intelligence operations for the answer economy” may be a useful second sentence, once the first one has done the orientation work. AI systems need this floor too. If the first page a system sees is built from soft outcomes, it may reach for a familiar category nearby. A managed program becomes consulting. A research operation becomes analytics. A visibility audit becomes SEO. These summaries can be plausible while still being commercially wrong. A service page that refuses to name the thing invites others to name it for you. ## Buyers read service pages to reduce risk Companies often write service pages as if buyers were looking for inspiration. In B2B, buyers are usually looking for risk reduction. They want to know whether the service fits their problem, whether the vendor understands their situation, what will happen after the first call, what they will receive, how much work it will require from their team, how long the engagement might take, how pricing or scope is determined, and what proof exists that the work is not invented for the landing page. A page can be elegant and still leave all of those questions unanswered. This is why the deliverable matters. A service that ends in “insights” is hard to buy. A service that ends in a visibility report, a source-trail map, a rewritten page set, a monthly monitoring brief, or an executive readout is easier to evaluate. The buyer can imagine forwarding it, defending it, budgeting for it, and deciding whether it is worth a conversation. Nielsen Norman Group has argued that B2B sites build trust when they disclose price information or at least help prospects understand representative pricing scenarios, because buyers need to feel educated rather than trapped in a sales process. [NN/g’s pricing research](https://www.nngroup.com/articles/show-price/) is older than the AI-search era, but the principle has become more important, not less. If a service is too complex for fixed pricing, the page can still explain what drives scope, what a small engagement looks like, or what kind of buyer should expect a larger program. Silence can feel flexible internally. To a buyer, it often feels evasive. ## The page should show its mechanism The weakest service pages describe benefits while hiding the mechanism. A company says it improves visibility, but not how visibility is examined. It says it provides research, but not who is researched or what is asked. It says it produces transparent reporting, but not what appears in the report. It says it uses human participation, but not what humans do. It says it helps with AI search, but not which systems are observed or how uncertainty is handled. The buyer is then forced to choose between trust and speculation. Many will choose neither. A service page does not need to publish the entire operating manual. It does need to show enough of the mechanism that the claim feels grounded. A page about an audit should explain the evidence it examines. A page about managed operations should explain the operating rhythm. A page about research should describe the kind of participant behavior being studied. A page about AI visibility should acknowledge that answer systems vary and that the work improves conditions rather than guaranteeing exact outputs. The mechanism is where credibility lives. A vendor that can describe how the work happens usually sounds more real than one that only describes what the client will feel afterward. This matters for AI readability as well. Google’s guidance for AI features says the same basic SEO and helpful-content principles apply to AI Overviews and AI Mode, and that there are no special AI text files or special schema.org structured data required for eligibility. Google emphasizes indexable, useful, textual content and alignment between structured data and visible page content. [Google Search Central’s AI features guidance](https://developers.google.com/search/docs/appearance/ai-features) is less magical than much of the GEO advice circulating online. The implication is plain. A service page should not try to game AI systems with decorative structure. It should contain the information a serious reader would need. ## Information scent matters more than page length A long service page can still be hard to use. A short page can be clear. Length is not the main issue. Scent is. Nielsen Norman Group uses “information scent” to describe the cues people use when deciding whether a link or page is likely to lead them toward the information they need. [The concept](https://www.nngroup.com/articles/information-scent/) is especially useful for service pages because buyers rarely read them linearly. They scan headings, follow labels, look for familiar nouns, and judge whether the page is worth more effort. A page with strong scent gives the buyer confidence that the answer is nearby. The heading names the service. The introduction names the buyer’s problem. The next section resembles a question the buyer actually has. A link to “Sample report” leads to a sample report, not a contact form. A section called “What’s included” explains the work, not just the benefits. A page with weak scent uses branded section titles that require decoding. “Signal Activation” might mean outreach, analytics, research, paid traffic, or something else. “Operational Layer” might be a service, a feature, or a philosophy. “Visibility Engine” might be a dashboard, a team, a workflow, or a metaphor that survived too many internal revisions. Weak scent creates a small humiliation for the buyer. They feel as if they should understand, but the page is not helping them. Many leave before they can articulate why. ## Proof should sit near the claim it supports B2B websites often separate claims from proof. The homepage says the company works with enterprise teams. The proof is hidden in a case study library. The service page promises transparent reporting. The sample report appears nowhere. The company says the process is research-led, but the methodology is buried in an old blog post. This separation weakens the page. Buyers experience doubt at the moment a claim appears, not three clicks later. If the page asks them to believe something important, the evidence should be close enough to answer the doubt while it is fresh. Proof does not always require a full case study. A short example, a sample output, a screenshot, a quote, a process detail, a data source, or a before-and-after explanation can be enough to make a claim feel less airy. The point is to turn adjectives into evidence. If the page says “transparent reporting,” show what is reported. If it says “real user interpretation,” show what kind of user observation is captured. If it says “AI visibility monitoring,” explain whether the work looks at mentions, citations, accuracy, sources, competitors, or prompt sets. If it says “multi-market,” show how market differences are handled. This is not only a trust issue. It is a source issue. A page with visible proof gives answer systems and third-party writers more specific material to reuse. A page made of claims gives them little except paraphrased marketing. ## Boundaries make the page more believable A service page becomes more credible when it says what the service does not do. This is uncomfortable for marketing teams because boundaries feel like lost sales. In practice, boundaries often reduce the wrong kind of interest and increase trust with the right buyer. A company that says it does not guarantee exact AI answers sounds more serious than one that implies it can control them. A research service that says its findings are qualitative and directional sounds more credible than one pretending five user sessions can predict an entire market. A managed program that explains what is outside scope prevents the buyer from imagining an unlimited custom engagement. Boundaries also help category clarity. If buyers might confuse the service with SEO, PR, analytics, consulting, or paid acquisition, the page should explain the relationship. Not defensively. Calmly. The buyer is already making the comparison; the page may as well make it more accurate. This is especially important for new categories. A service page in an established category can rely on buyer familiarity. A service page in an emerging category has to build the category while selling the service. That requires more patience and less slogan work than most pages contain. ## A service page should survive being summarized One useful test is to paste the page into a neutral summarizer or ask someone who does not know the company to describe it after two minutes. If the summary is wrong, vague, or oddly generic, the page probably has a source-document problem. The page should not read like a database entry. It can still have voice, and it should still carry the company’s point of view. The basic commercial facts, though, need to survive compression: what the service is, who it is for, what problem it addresses, what work happens, what the buyer receives, and what proof supports it. A service page that survives summarization is useful to everyone. The buyer can forward it. The sales team can rely on it. A founder can use it as public reference. Search systems can classify it. AI systems have less room to flatten it into the wrong category. The service page used to be one stop in the funnel. Now it is also part of the public record. That makes it less disposable than a landing page and more demanding than a brochure. It has to sell, but first it has to testify. --- ## The Sales Deck Is Where Public Evidence Goes to Die Source: https://zartrace.com/en/research/sales-deck-public-evidence-goes-to-die/ > Many B2B companies already have the proof buyers need. The problem is that it is trapped in private material. A prospect asks for an example. The account executive smiles because this is the easy part. There is a slide for it: before-and-after metrics, a short customer story, a screenshot of the reporting interface, a quote from a team lead, a neat diagram of the implementation process. The deck has been refined over months. It answers the question well. The website does not. On the public service page, the same claim appears in a sentence so broad it could belong to any company in the category. “We provide transparent reporting and measurable outcomes.” No screenshot. No sample output. No explanation of what gets measured. No messy detail about how long the first cycle takes. No sign that the company has already solved the problem for someone like the buyer. This is a composite scenario, but the pattern is common. B2B teams often keep their strongest proof inside the sales process. They show the evidence after the buyer has already agreed to talk. They forget that the buyer wanted the evidence before deciding whether the conversation was worth having. The sales deck becomes a vault. Useful material enters. Almost none of it escapes. ## Private proof does not help the buyer who is still deciding whether to speak There are good reasons to keep some evidence private. Customer names may be sensitive. Metrics may be confidential. Screenshots may show internal workflows. Legal teams may prefer caution. Enterprise sales often requires context that cannot be flattened into a public page. Still, many companies are more private than they need to be. They hide not only confidential details, but basic explanatory evidence: what an audit contains, what a report looks like, what a project timeline feels like, what a client receives, how the team handles uncertainty, what kind of change is realistic after thirty days, and what the service explicitly does not guarantee. A self-directed buyer cannot evaluate what they cannot see. Gartner reported in March 2026 that 67% of B2B buyers prefer a rep-free experience, and 45% of surveyed buyers had used AI during a recent purchase. [Gartner release](https://www.gartner.com/en/newsroom/press-releases/2026-03-09-gartner-sales-survey-finds-67-percent-of-b2b-buyers-prefer-a-rep-free-experience) Those numbers should not be stretched into a universal law, but they describe a real pressure: buyers increasingly expect to learn before a vendor conversation. When the public site withholds the useful parts, the buyer has to choose between booking a call too early or moving to a vendor that explains more. Many will choose the second path. They will not announce that your proof arrived too late. They will simply disappear into the company whose proof was easier to inspect. ## AI systems cannot cite your private confidence The same problem affects AI visibility. An answer system cannot summarize a slide it cannot access. It cannot cite a private deck. It cannot infer a method from a sales narrative that only exists in calls. This matters because AI-mediated discovery often asks for evidence-rich answers. A buyer may ask which vendors are credible, what a service includes, how one provider differs from another, or whether a company has proof of outcomes. If the public web contains only vague claims, the system has little to use. A competitor with public case studies, sample reports, clear service pages, and third-party profiles becomes easier to describe. The competitor may not have better work. It may simply have more public evidence. This is one of the stranger unfairnesses of AI search. A company with a cautious but excellent private sales process can look weaker than a company with a mediocre product and better public artifacts. The answer environment is not reading your internal Slack threads, your QBR deck, your implementation notes, or the explanation your best salesperson gives on the second call. It is reading what the public web can provide. A private truth has limited value in public discovery. ## A deck has a narrator. A public page does not. Sales decks are built to be spoken over. A vague diagram can work if the account executive knows where to pause. A screenshot can be useful if someone explains what matters inside it. A metric can be persuasive if the seller knows the messy background: why the baseline was unreliable, why the first month looked worse before it looked better, why the customer cared less about the headline number than about the internal confidence it created. Public evidence has no narrator. It has to carry more of its own context. This is why simply copying slides into a website rarely works. The slide was designed as a prompt for conversation, not as a self-contained source. Once it becomes public material, the missing narration has to be written back in. What is being shown? Why does it matter? Which part is typical, and which part is specific to that client? What should the reader not overgeneralize? That last question is especially important. Public evidence becomes more believable when it refuses to pretend that every example is universal. A sample report can say that the structure is typical while the findings are illustrative. A case study can say that the timeline depended on the client’s data access. A pricing example can say which assumptions would change the scope. This kind of caveat does not weaken the proof. It makes the proof feel handled by adults. ## Evidence does not have to reveal everything to be useful Some teams avoid public proof because they imagine the only alternative is oversharing. They think a case study must name the customer, reveal the number, show the screenshot, describe the entire workflow, and survive procurement review. If that is the standard, many teams will publish nothing. There is a middle ground. A company can show an anonymized report excerpt. It can describe the shape of an engagement without naming the customer. It can publish a composite example and label it clearly. It can show the table of contents of an audit without showing client data. It can explain how findings are categorized. It can describe common patterns seen across projects. It can publish a sanitized screenshot. It can give a realistic range instead of a precise number. It can explain what a client receives in the first week, the first month, and the first review cycle. None of this replaces a full sales conversation. It prepares the buyer for one. Nielsen Norman Group’s work on web trustworthiness identifies up-front disclosure and comprehensive, current content as durable credibility factors. [NN/g on trustworthy design](https://www.nngroup.com/articles/trustworthy-design/) In B2B, disclosure often means showing enough of the work that the buyer does not have to treat the vendor’s claim as a black box. A sample output can do more for trust than another sentence about trust. ## Sales teams often know what the website should say There is a small tragedy in many B2B organizations: the sales team has already learned the buyer’s real questions, but the website still answers the marketing team’s imagined ones. A good sales deck evolves under pressure. If a slide does not work, prospects ask again. If a claim creates confusion, the account executive adds an example. If a diagram saves ten minutes on every call, it stays. If a screenshot makes the buyer relax, the team uses it more often. The deck becomes a rough empirical record of what buyers need to see. The website often evolves more slowly. It goes through brand reviews, design cycles, SEO briefs, and internal politics. It may sound more polished than the deck while being less useful. That gap is worth studying. The best public evidence often begins as a sales artifact that has proven itself in conversation. A repeated objection becomes a public explanation. A useful slide becomes a service-page section. A report walkthrough becomes an article. A common clarification becomes a diagram. A hidden proof point becomes a public pattern. Dumping the sales deck onto the site would be a mistake. Decks are built for live narration, and many are unreadable without the person presenting them. The better move is to identify which pieces of private proof repeatedly create understanding, then translate those pieces into public source material. ## The best evidence is usually a little imperfect Many companies delay publishing proof because it is not polished enough. The customer quote is short. The metric is directional. The report screenshot has too much internal terminology. The implementation story is not dramatic. The client would rather remain unnamed. The before-and-after is real but not cinematic. That imperfection can be useful. Buyers often trust evidence that feels proportionate. A case study that claims total transformation in six weeks may sound less believable than an anonymized example showing a modest but specific improvement. A sample report with a few rough edges may feel more real than a glossy diagram. A limitation stated plainly can make the surrounding claim more credible. AI systems also benefit from specificity. A page that says “we provide strategic insights” gives little to summarize. A page that explains the structure of the report, the kinds of evidence included, the review cadence, and the conditions under which findings are uncertain becomes more usable source material. Good evidence does not need to be theatrical. It needs to be inspectable. ## The website should not make sales repeat the basics When public pages are thin, sales spends too much time doing orientation. The first call becomes a guided tour through information that could have been public. The account executive explains the category, the process, the output, the difference from alternatives, the reason the service costs what it costs, and the proof that the company is real. Some of that conversation is valuable. Too much of it is repair. A better public evidence layer changes the sales call. The buyer arrives with a more accurate picture. The call can focus on fit, constraints, timing, and internal priorities rather than basic explanation. The salesperson is no longer rescuing the website from vagueness. This also improves AI-mediated discovery. A public page that answers the basic questions can become material for summaries, citations, comparisons, and buyer research. The same evidence that helps a cautious buyer helps a system describe the company with fewer guesses. The sales deck still matters. It just should not be the first place the company becomes understandable. ## Public evidence is a form of respect There is a cultural habit in B2B of treating detailed explanation as something the buyer earns by booking a call. That habit is increasingly out of step with how buyers behave. A buyer who wants to research independently is not being difficult. They are trying to avoid wasting time, political capital, and budget attention. Giving them useful public evidence is a form of respect. It says the company understands the decision is not casual. It says the buyer can inspect the work before surrendering their calendar. It says the vendor is not relying entirely on controlled conversation to make the offer look credible. The companies that adapt will not publish everything. They will publish enough. Enough to make the service legible. Enough to make the claim checkable. Enough to give AI systems and buyers a public version of the proof that used to die inside the deck. The deck can still close the deal. But the evidence should start working before the meeting invite is sent. --- ## Robots.txt Is Becoming a Brand Visibility Decision Source: https://zartrace.com/en/research/robots-txt-brand-visibility-decision/ > The question is no longer simply whether bots may crawl the site. It is which parts of the public brand you want answer systems to be able to see. A publisher reads another story about AI companies scraping the web and tells the technical team to block everything that looks like an AI crawler. The instruction is emotionally understandable. The company produces expensive content. It does not want that work absorbed into systems that may summarize it without sending traffic back. A B2B software company hears the same story and reacts differently. Its marketing team wants ChatGPT and Perplexity to find its service pages, cite its case studies, and stop describing the company from old third-party profiles. The CTO asks a practical question: which bots should be allowed, which should be blocked, and what is the difference between search retrieval and model training? For years, robots.txt felt like infrastructure plumbing. It still is. But in AI-era discovery, crawler policy is also becoming a brand visibility decision. It affects whether public pages can be found, retrieved, summarized, cited, and used as current evidence. It also affects how much control a company keeps over intellectual property, paid content, sensitive material, and original research. The issue is not clean. The same setting can look like protection to one team and self-sabotage to another. ## AI crawlers are not all doing the same job The first mistake is treating all AI bots as one creature. OpenAI’s crawler documentation says it uses different web crawlers and user agents for different purposes, including OAI-SearchBot and GPTBot, and that robots.txt tags can help webmasters manage how their sites and content work with AI. [OpenAI crawler docs](https://developers.openai.com/api/docs/bots) OpenAI’s publisher FAQ says sites that allow OAI-SearchBot can track referral traffic from ChatGPT using analytics tools, with ChatGPT adding a `utm_source=chatgpt.com` parameter in referral URLs. [OpenAI publisher FAQ](https://help.openai.com/en/articles/12627856-publishers-and-developers-faq) The technical distinction matters. A crawler used for search retrieval is not the same as a crawler used for training. A user-triggered fetch is not the same as bulk crawling. A bot that helps answer a live question creates different tradeoffs from a bot collecting material for model development. Nontechnical teams often collapse these distinctions because the public conversation around AI scraping is noisy and adversarial. That can lead to blunt policies: allow everything or block everything. Both choices can be wrong depending on the business model. A news publisher, a documentation site, a B2B service company, a product review site, and a private community do not have the same incentives. A company selling visibility and trust may want its public explanatory pages available to answer systems. A company monetizing original content through subscriptions may want narrower access. A company with sensitive customer material should probably be more conservative than either. Crawler policy should begin with the business model, not the mood of the last article the leadership team read. ## Blocking can protect content and weaken the current story Blocking AI crawlers may be the right choice for some content. It can protect paid work, reduce unwanted scraping, support licensing strategy, and keep sensitive or high-value material from being used in ways the company does not accept. But for brands trying to be understood in AI-mediated discovery, blocking can create an unintended side effect. If answer systems cannot access the current website, they may rely more heavily on whatever else remains available: old directories, third-party summaries, competitor pages, review profiles, cached descriptions, or public sources the company does not control. This can leave the company in a strange position. The official site is protected, but the public story is being assembled from weaker witnesses. A B2B company that blocks access to its service pages may later wonder why ChatGPT describes it from a stale profile. A local business that hides useful public information may leave answer systems with only review snippets. A SaaS company that protects its best explanatory content may lose the opportunity to have that content cited in a buyer’s early research. The right answer is not always to open everything. The point is that blocking has a visibility cost. The cost may be worth paying. It should still be recognized. ## The page you block may be the page that corrects the record The crawler conversation often treats content as if every page has the same commercial role. That is rarely true. Some pages are assets to protect. Others are corrections to a public misunderstanding. A service page that explains the current offer, an About page that clarifies the company’s category, a case study that shows the work, or a pricing page that reduces uncertainty may be doing more than attracting traffic. It may be repairing the source trail. This becomes obvious when the current website is the only place where the company is described accurately. Imagine a company that has just moved from software to a managed service. Old directories still call it a tool. Review profiles still discuss features from the earlier product. Competitor pages use the old category because it makes comparison easier. If the company blocks legitimate search and answer systems from the new explanatory pages, it may protect the very pages needed to correct the public record. That does not mean every page should be open. It means crawler policy should recognize the difference between proprietary depth and public correction. A public source-of-truth page has a different job from a gated research report. Blocking both with the same rule may feel clean technically and create a mess commercially. ## Allowing access is not the same as losing control Some teams treat crawler access as if it were a binary surrender. Either the public pages are available to AI systems, or the company has lost control. The reality is more granular. Robots.txt can distinguish between crawlers. Page-level controls can shape whether content appears in search results and snippets. Google’s documentation for AI features says publishers can use controls such as `nosnippet`, `data-nosnippet`, `max-snippet`, or `noindex` to manage preview content and indexing in Search, although those same controls affect traditional search displays as well. [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features) This is not a magic governance layer. It is a set of imperfect controls in a fast-changing environment. Still, it means the conversation can be more precise than “block AI” or “allow AI.” A company might allow access to public service pages, case studies, help documentation, and thought leadership, while restricting gated research, private customer portals, internal documentation, or content whose business value depends on direct access. It might allow search-related bots while blocking training-related bots where the platform supports that distinction. It might keep short public summaries accessible while requiring login for deeper proprietary material. The brand question is: which pages do we want answer systems to be able to use as the current public version of us? That question belongs in the same room as marketing, legal, security, product, and leadership. It should not be left entirely to a default CDN setting. ## Cloudflare turned a quiet setting into a strategic interface The crawler debate has become visible because infrastructure companies are turning it into product controls. Cloudflare’s AI Crawl Control describes the ability to block, allow, or charge certain AI crawlers, and says it gives site owners granular insight into AI crawler activity on their domains. [Cloudflare AI Crawl Control](https://www.cloudflare.com/ai-crawl-control/) Cloudflare has also written about the growth of AI bot traffic and the tension between crawling, traffic, and compensation. Its reporting and product moves are not neutral academic evidence; Cloudflare has its own commercial position in the debate. But the fact that crawler management has become a visible product category tells us something. This is no longer an obscure technical footnote. It is becoming part of how organizations manage the public web. For brand discoverability, the practical value is not only blocking. It is observability. Many teams do not know which AI-related crawlers visit their sites, which pages they request, whether important pages are accessible, or whether the official source of truth is actually visible to systems that might summarize the company. A crawler log can be boring. It can also answer a brand question: are the systems that might describe us able to reach the pages that describe us best? ## The policy should vary by content type A single AI crawler policy for an entire domain may be too crude. A public service page exists to explain the business. A help article exists to solve a user problem. A case study exists to demonstrate proof. A pricing page exists to reduce uncertainty. A research report may exist partly as a lead magnet. A private customer document exists for a different reason. A paywalled article may be the business itself. These pages have different exposure logic. For a company whose main problem is being misdescribed, the public explanatory pages should probably be easy for legitimate search and answer systems to access. For a publisher whose revenue depends on subscriptions, the calculus is different. For a company with sensitive vertical expertise, some material may need public summaries and protected depth. For a company whose current brand is being drowned by old third-party sources, access to the current site may be part of the correction. The worst policy is the one nobody meant to set. A blanket block added during a panic. A default allow left in place because nobody checked. A robots file copied from another site. A CDN setting changed by security without marketing knowing. A marketing team demanding citations from systems the site has quietly blocked. Crawler policy is now cross-functional because the tradeoffs are cross-functional. ## Visibility without governance is naive. Governance without visibility is brittle. The crawler debate often gets stuck between two camps. One wants maximum openness because visibility matters. The other wants maximum restriction because content has value and AI companies have not earned trust. Both instincts are understandable. Neither is sufficient by itself. A company that allows everything may lose control over material that should have been protected. A company that blocks everything may make itself harder to describe accurately in the very systems buyers use to research it. The more useful position is deliberate exposure: decide which parts of the public brand should be visible, which parts should be protected, and which bots or contexts are acceptable. This requires maintenance. Platforms change. New crawlers appear. Documentation updates. AI search becomes more entangled with traditional search. Referral patterns shift. Legal and licensing norms remain unsettled. A policy set once in 2024 may not fit 2026. The web used to reward being crawlable. AI search has made the question more political, more commercial, and more brand-sensitive. Robots.txt still looks like a small text file. It is increasingly a public-access policy for the machine readers that stand between buyers and your brand. --- ## Reviews Are Becoming Machine-Readable Reputation Source: https://zartrace.com/en/research/reviews-are-machine-readable-reputation/ > A review used to sit on a profile page waiting for a buyer. Now it may be summarized before the buyer ever opens the profile. A software company has forty-seven reviews on a major platform. The average rating is good. Nobody is worried. The customer marketing team checks the profile before renewal season, responds to the occasional complaint, and uses a few quotes in sales material. Reviews are treated as social proof: useful, but separate from the main website and the content strategy. Then an AI answer summarizes the company as “easy to deploy but limited for larger teams.” The phrase comes from somewhere. Not one review exactly, but a pattern. A few older customers mentioned quick setup. Two reviews from 2023 complained about enterprise reporting. A competitor comparison repeated the same limitation. The company fixed the reporting issue last year, but the old review language still sits there, public and legible. The AI answer is not entirely wrong. It is worse: it is outdated in a way that sounds current. This is a recurrent pattern in AI-mediated discovery. Reviews are no longer only destination pages that buyers visit after narrowing a shortlist. They are becoming machine-readable reputation: text that can be searched, summarized, cited, aggregated, and folded into recommendations before the buyer inspects the original source. The review profile has become part of the brand’s source trail. ## Reviews are written for humans, but machines can read the residue A review is usually messy in a human way. It mixes product feedback, implementation emotion, team frustration, pricing anxiety, support experience, and sometimes a small complaint that has nothing to do with the vendor. A customer writes after a good onboarding, or after a renewal dispute, or after a feature request was denied, or because the customer success manager asked at the right moment. For a human buyer, that mess is part of the value. Reviews feel less controlled than vendor copy. They contain hesitation, odd phrasing, praise that is too specific to be invented easily, and complaints that help the buyer judge tradeoffs. For AI systems, the same mess becomes analyzable text. OpenAI’s documentation for ChatGPT shopping says ChatGPT may display model-generated review summaries based on reviews from public websites, intended to highlight common likes and dislikes, while also noting that reviews and ratings are not verified by OpenAI. [OpenAI shopping docs](https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search) That page is about shopping, not B2B software procurement. Still, it makes the mechanism visible: public review text can become summarized review intelligence. G2’s April 2026 software-buyer research reported that review site citations were the top signal making buyers trust an AI chatbot’s recommendation. The same release said 51% of B2B software buyers now start research with an AI chatbot more often than with Google. [G2 research release](https://www.prnewswire.com/news-releases/new-g2-research-half-of-b2b-software-buyers-now-start-their-research-with-ai-chatbots-302742807.html) Survey data should be read with attention to category and methodology, but the direction is clear enough. Reviews are moving upstream. They can shape the shortlist before the review site visit. ## The words matter more than the stars Ratings are easy to notice because they are numerical. A 4.7 feels better than a 4.1. Review count matters too. A profile with two reviews feels thinner than a profile with two hundred. But in AI-mediated discovery, the language of reviews may matter even more than teams expect. The words carry category signals. Customers describe what they believe the product does. They name use cases. They compare alternatives. They complain about missing features. They praise support, speed, accuracy, flexibility, onboarding, price, or integration. They say who the product is good for and who should be careful. Over time, those phrases form a rough public taxonomy of the company. The vendor may describe itself as an enterprise platform, but if reviews repeatedly say “great for small teams,” that phrase has weight. The vendor may position around strategic visibility, but if customers mostly praise fast reports, the public reputation may narrow around execution speed. The vendor may have fixed a weakness, but if the old weakness is still repeated in review text, the public memory may lag. A review profile is not only a scorecard. It is a body of language about the company. That language can help or hurt depending on whether it matches the current business. ## Old reviews can preserve old products Reviews often outlive the version of the company they describe. A customer reviews a product in 2022. The company changes pricing in 2023. It adds a managed service in 2024. It shifts upmarket in 2025. By 2026, the review is still public, still indexed, still emotionally convincing, and still attached to the current brand. The customer was telling the truth at the time. The truth has expired. Expired truth is harder to manage than falsehood. A false review can sometimes be disputed. A fake review can be flagged. A policy violation may be removed. But a real review about an older version of the company often remains because it is legitimate. It is not wrong historically. It is wrong as a current summary. This is where review strategy becomes more than collecting positive comments. The company needs enough recent, specific review language to balance older public memory. If the product is now enterprise-ready, recent customers should be able to say what changed. If the service model has matured, the public record needs current descriptions of that model. If implementation is no longer painful, old complaints cannot be the only detailed text available. The past does not need to be erased. It does need to stop being the easiest material to summarize. ## AI summaries flatten nuance, so the source needs more nuance A buyer reading reviews manually can notice distribution. They can see that one complaint is old, that another comes from a different use case, that a glowing review may be from a small customer, or that a criticism is relevant only to a feature they do not need. An AI summary may flatten that nuance. It may say “users praise ease of use but cite limitations in reporting” without clearly separating 2023 from 2026, small-team users from enterprise users, or implementation complaints from product complaints. This is not always the system behaving badly. Summarization compresses. Compression loses edges. Brands cannot fully control that compression, but they can improve the source material. Recent reviews can be more specific. Responses to reviews can clarify what changed. Public changelogs, case studies, help docs, and service pages can provide updated context. Review profiles can be managed so that the current customer base is represented, not only the earliest or loudest users. Google’s Business Profile documentation reminds businesses that reviews appear next to Business Profiles in Search and Maps and can give potential customers helpful information. [Google Business Profile docs](https://support.google.com/business/answer/3474122) That statement is local-search oriented, but it captures the older version of the function: reviews help customers decide. The newer layer is that reviews may also help machines summarize how customers decide. The review is no longer only read. It is processed. ## Fake reviews are more dangerous when reviews become data The obvious temptation is to treat review visibility as another surface to manipulate. That is a bad idea ethically, and increasingly a bad idea operationally. Google’s Maps user-generated content policy says contributions should reflect a genuine experience at a place or business, and that fake engagement is not allowed. [Google contribution policy](https://support.google.com/contributionpolicy/answer/7400114) OpenAI’s shopping documentation explicitly says reviews and ratings shown in ChatGPT are not verified by OpenAI. [OpenAI shopping docs](https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search) That creates a messy environment. Public reviews are influential, but not perfectly verified. They can be summarized, trusted, doubted, cited, misread, or abused. Fake reviews may help a surface look better briefly, but they also pollute the evidence layer that buyers and systems depend on. Once reputation becomes machine-readable, low-quality manipulation does not merely risk a platform penalty. It risks teaching downstream systems a public story that may later be distrusted or contradicted. A serious company should not want a perfect review profile. It should want a credible one. Credibility usually includes some criticism. A profile with only generic praise can feel less useful than one with specific strengths, specific limitations, and thoughtful vendor responses. Buyers are not looking for a company without tradeoffs. They are looking for tradeoffs they can understand. ## Responding to reviews is part of source-trail maintenance A review response used to be seen as customer-service etiquette. In AI-era discovery, it can also become public context. A thoughtful response can clarify that a feature has changed, that a limitation applies only to one plan, that a support issue was addressed, or that the company now handles onboarding differently. It cannot rewrite the customer’s experience, and it should not try. But it can add a current layer around an older claim. The best responses are not defensive. They are factual, specific, and calm. They recognize the customer’s experience, explain what is true now if something has changed, and avoid turning the review profile into a PR argument. This matters because review pages are read by cautious buyers. They may also be read by systems looking for public context. A profile where the company ignores criticism feels different from one where the company responds with useful specificity. The second profile gives the buyer more material for judgment. Reputation is not the absence of negative text. It is the presence of enough honest context to interpret that text. ## Reviews now belong to more than customer marketing Review management often sits in an awkward organizational corner. Customer marketing wants quotes. Customer success wants happy customers to be recognized. Sales wants proof. Product may read reviews for feedback. SEO may care about profile visibility. Brand may care only when something embarrassing appears. AI-mediated discovery makes that division less sustainable. Reviews affect how buyers interpret the company, how AI systems may summarize public sentiment, how competitors are compared, and how current the brand’s public memory feels. They are part of the discoverability system. No company needs to turn reviews into a paranoid monitoring operation. Reviews should simply be read as public language about the business, not just as ratings. The company should know which words customers use, which old complaints still shape perception, which strengths are repeated, which segments are represented, and where the current offer is missing from the review record. A review profile is a mirror, but it is not a passive mirror. It can be quoted, summarized, cited, and carried into places the company does not control. The question is no longer only whether reviews make the company look good. It is whether they still describe the company that exists now. --- ## The Difference Between Real User Research and Fake Engagement Signals Source: https://zartrace.com/en/research/real-user-research-vs-fake-engagement-signals/ > Real research leaves the company with a sharper view of how people understand it. Fake engagement leaves the company with a busier chart. A founder hears the phrase “human signals” and becomes suspicious before the sentence is finished. The suspicion is earned. Digital marketing has trained serious people to distrust words like engagement, signals, behavioral data, and visibility. Too often, they have been used as softer names for low-quality traffic, fake reviews, automated clicks, synthetic accounts, or dashboards full of movement that nobody can interpret. The surface looks active. The business learns almost nothing. Now imagine two vendors describing work that sounds similar from a distance. One recruits real people to review a website, compare it with competitors, describe what they think the company does, identify confusing claims, and explain what proof they would need before contacting sales. The output is messy but useful: quotes, observations, repeated misunderstandings, and recommended changes to the site and public profiles. The other sends unexplained traffic to the same site, produces a chart showing higher engagement, and says the “signals” improved. Both can use the language of human behavior. Only one is research. This distinction matters because the market around AI visibility, GEO, user signals, and answer-engine optimization is becoming noisy. Some of the work is legitimate: studying how people and AI systems discover, interpret, compare, and trust a company. Some of it is old manipulation wearing new clothes. The difference is not philosophical. It shows up in whether the company becomes smarter. ## Fake engagement creates numbers without memory Fake engagement is attractive because it moves. A traffic graph rises. A session count appears. A heatmap gets warmer. A click-through rate changes. The vendor has something to report. The client has something to show internally. Everyone can pretend the surface motion is evidence of progress. But the business cannot answer the questions that matter. Did real buyers understand the offer? Did they compare the company with the right alternatives? Did the homepage explain the category? Did the service page answer the question that normally appears before a sales call? Did the public sources around the brand support the same story? Did users trust the proof, or did they merely fail to object because nobody asked them? Fake engagement is sterile. It cannot explain itself. It also damages the company’s ability to learn. If analytics are polluted with artificial behavior, the team has less confidence in its own data. If fake reviews enter the public record, the trust layer becomes brittle. If a vendor refuses to explain where traffic came from, the client is not buying visibility; it is renting opacity. A serious company should not want unexplained movement. It should want interpretable evidence. ## Real user research has texture Real user research is slower and less theatrical. It produces fewer dramatic charts. It also produces sentences a team remembers. “I thought this was software until the pricing page made it sound like consulting.” “I understand the problem, but I do not know what I would actually buy.” “The homepage feels credible, but I would need to see an example report.” “I would compare this with an SEO agency unless the page explains why that is wrong.” These comments are not perfect data. They may come from small samples. They require judgment. They can be overread if handled carelessly. But they have texture because they show how a person is assembling the company from the material available. That texture is valuable because B2B discoverability often fails in small interpretive gaps. The buyer understands the topic but not the offer. They trust the founder but not the process. They like the category but cannot see the deliverable. They believe the claim but cannot find enough proof to share internally. Real research does not pretend these gaps are ranking factors. It treats them as friction in the buyer’s understanding. ## Small studies can be useful if the claim stays small There is a common objection to qualitative research: the sample is too small. Sometimes that objection is fair. A handful of users cannot estimate market size, predict conversion rates, or prove that every buyer will react the same way. But small studies can be extremely useful when the question is diagnostic. Nielsen Norman Group’s usability-testing guidance has long argued that small qualitative tests can uncover many common usability problems when they involve representative users performing realistic tasks. [NN/g’s Usability Testing 101](https://www.nngroup.com/articles/usability-testing-101/) recommends five participants for a typical qualitative usability study of a single user group. Five is not a magic number for all research; the underlying point is that repeated observation can quickly reveal the most obvious problems. The same logic applies to website interpretation. If five reasonably matched users read a service page and four cannot tell whether the company sells software or a managed service, the team does not need a thousand-person survey to suspect a clarity problem. If several users ask for the same missing proof, that is not a market forecast. It is a design and messaging problem with enough recurrence to justify action. The discipline is in the claim. A small study can say, “This pattern appeared repeatedly and should be tested or corrected.” It should not say, “The market has spoken.” Fake engagement rarely makes this distinction. It wants the authority of data without the burden of interpretation. ## The method should be boringly explainable Responsible human-signal work should be easy to describe. Who participated? What were they asked to do? What did they see? What did they say? What patterns appeared? What was uncertain? What changed because of the finding? If a provider cannot answer those questions, the work is not mature enough. If the method has to remain secret for the service to sound valuable, that is a warning sign. There are legitimate reasons to keep some operational detail private, especially around participant sourcing or anti-fraud controls. The structure of the work should still be understandable. Research does not need theatrical secrecy. It needs enough transparency that the client can trust the output. The difference between a study and a stunt is often documentation. A study records the task, context, participant type, observed behavior, and interpretation. A stunt records a metric and asks the client to believe it. ## Good research marks uncertainty One of the clearest differences between real research and fake engagement is how each handles uncertainty. Fake engagement wants to sound conclusive. It says signals improved. It says behavior changed. It says the campaign worked. The language is smooth because the mechanism is hidden. Real research is more careful. It may say that several participants misunderstood the service model. It may say that the sample was small, but the pattern is worth investigating. It may say that users found the homepage credible but could not name the deliverable. It may say that a competitor was easier to understand because its category label was more familiar. It may say that the finding should be checked again after the page changes. This caution is not weakness. It is what makes the work usable. A small qualitative study should not pretend to be a market-size estimate. A prompt test should not pretend to be a permanent AI ranking. A perception audit should not claim to predict every buyer’s reaction. Responsible work separates observation from interpretation. That distinction is especially important in AI visibility. AI systems vary by prompt, source access, model behavior, location, and time. A single answer can be interesting without being definitive. A repeated pattern across prompts and systems is more useful. The work should say which is which. ## Research changes the artifact The strongest sign that user research is real is that it changes something concrete. A homepage definition gets rewritten. A service page adds an example deliverable. A comparison article explains the wrong competitor frame. A pricing page adds scope context. A directory profile is corrected. A case study moves closer to the claim it supports. An FAQ answer becomes more honest. A public profile stops using old category language. Fake engagement usually leaves the artifact untouched. It treats the website as a stage for activity rather than a thing to be improved. Real research treats the website, source trail, and discovery surfaces as editable objects. The point is not to watch users struggle forever. The point is to learn where they struggle, then reduce the struggle. This is where human-signal work connects to AI-era discoverability. The same materials that help people understand a company often help answer systems describe it more accurately: plain category language, current profiles, visible proof, concrete service descriptions, and fewer contradictions across the public web. The research may begin with humans. The improvements travel further. ## Some signals are legitimate because they are ordinary There is a quiet irony in this category. The legitimate signals are often the least exotic ones. People reading a service page. People comparing two vendors. People checking reviews. People asking whether a claim is believable. People trying to explain what a company does after two minutes on the site. People noticing that the LinkedIn description and homepage do not match. These behaviors are plain, almost dull. They are useful because they resemble how buyers actually evaluate companies. The questionable signals tend to be the ones that detach behavior from intention: clicks without context, sessions without participants, reviews without customers, engagement without explanation. Activity that exists to imitate the residue of interest rather than generate understanding. A real human signal has a human reason behind it. Someone was trying to understand, compare, verify, decide, or explain. Once that reason disappears, the signal becomes decorative. ## The legal and ethical line is getting less abstract Fake reviews are the clearest example of why this distinction matters beyond style or methodology. In 2024, the U.S. Federal Trade Commission announced a final rule banning the sale or purchase of fake reviews and testimonials and allowing civil penalties against knowing violators. [The FTC announcement](https://www.ftc.gov/news-events/news/press-releases/2024/08/federal-trade-commission-announces-final-rule-banning-fake-reviews-testimonials) described fake reviews as a market problem serious enough to require explicit enforcement. The rule is about consumer reviews rather than every possible form of engagement manipulation. Still, it signals the broader direction: fabricated trust is becoming harder to excuse as marketing. Even when a practice is not legally identical to fake reviews, the practical logic is similar. If the activity creates the appearance of human validation without real human judgment, it weakens the evidence layer. It makes the company look more active while making the public record less trustworthy. For reputation-sensitive B2B companies, that tradeoff is usually irrational. The short-term metric is not worth the long-term doubt. ## The useful question When evaluating any “signal” work, the useful question is not whether humans were technically involved. Humans can be involved in useless or manipulative activity too. The better question is what the company will understand after the work that it did not understand before. If the answer is only that traffic went up, the work is thin. If the answer is that users misunderstood the service model, competitors were easier to explain, external profiles contradicted the homepage, buyers needed proof that was missing, or AI systems repeated an old category because the public trail was stale, then the work has substance. Real user research leaves the company with a sharper view of itself. Fake engagement leaves the company with a busier chart. --- ## A Prompt Audit Is a Lens, Not a Map Source: https://zartrace.com/en/research/prompt-audit-is-a-lens-not-a-map/ > Asking AI systems about your brand can reveal useful blind spots, but a prompt set is a way of looking, not a complete model of the market. A marketing team runs its first AI visibility audit on a Friday afternoon. Someone opens ChatGPT, someone else opens Perplexity, and a third person keeps a spreadsheet. They ask twenty questions that sound reasonable: what the company does, who its competitors are, what alternatives exist, which vendors help with the category, whether the brand is credible. A few answers are encouraging. A few are irritating. One answer names a competitor that nobody internally takes seriously. Another describes the company with language from an old product page. The team screenshots the strangest results and posts them in Slack. By Monday, the screenshots have become evidence. The founder wants to know why the company “ranks below” a competitor. Sales wants to use the good answer in a deck. Marketing wants to rewrite pages around the prompts where the brand was absent. Someone asks whether there is now a ChatGPT ranking report the way there used to be a keyword ranking report. This is where AI visibility work can go wrong. The audit was useful. The interpretation was too heavy. A prompt audit is not a map of the market. It is a lens. It shows how a small set of questions, asked in a particular way, through particular systems, at a particular moment, can cause the public web and model behavior to assemble a version of the brand. That version may be revealing. It may also be unstable, incomplete, and overly dependent on the wording of the question. The mistake is treating the answer as if it were a fixed position. ## Prompts are not keywords with a new costume It is tempting to think about prompts the way marketers learned to think about keywords. A keyword had search volume, intent, ranking position, click-through rate, and a page that could be optimized. The model was imperfect, but it gave teams a stable working fiction: this query exists in the market, this page ranks here, this competitor ranks there. Prompts do not behave as neatly. A buyer might ask an AI system a long, messy question that contains context, constraints, preference, uncertainty, and memory from earlier turns. They might ask for “good options for a mid-market team that already uses HubSpot and does not want another reporting dashboard.” They might ask for alternatives to a competitor but exclude agencies. They might ask the same thing twice with different phrasing and get a different emphasis. They might ask the model to explain a vendor to a CFO, then to a technical lead, then to a procurement team. A keyword is usually a fragment. A prompt is often a situation. That difference matters because the answer system is not merely matching a query to documents. It is interpreting the task. ChatGPT Search can decide to search the web depending on the user’s question and can return answers with links to web sources. [OpenAI’s documentation](https://help.openai.com/en/articles/9237897-chatgpt-search) also says ChatGPT may rewrite a user’s prompt into one or more targeted queries sent to search providers. Google describes a related pattern in AI Overviews and AI Mode: “query fan-out,” where multiple related searches may be issued across subtopics and data sources to develop a response. [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features) gives this as part of how these AI features can assemble broader answers. The prompt the user typed is therefore not always the search the system performed. It is the front door to a small investigation. A brand can be absent from the visible answer for reasons that have little to do with one clean “ranking.” The system may have expanded the question toward a subcategory where the brand has weak evidence. It may have chosen sources where competitors are better represented. It may have interpreted the prompt as asking for software when the company sells a managed service. It may have used a public list that omits the brand. It may have decided the answer needed general education rather than vendor recommendations. The output is still useful. It is simply not a ranking table. ## The shape of the question changes the company that appears A small wording change can expose a large positioning problem. Consider a composite example from B2B service categories. A company helps enterprise brands audit and improve how AI systems describe them. When the prompt asks for “AI visibility audit providers,” the company appears occasionally. When the prompt asks for “GEO agencies,” it disappears behind SEO firms. When the prompt asks for “brand perception research for AI search,” it appears again, but the answer places it next to market research vendors. When the prompt asks for “tools to track ChatGPT mentions,” it vanishes, which may be correct because the company is not really a software tool. None of these answers is the whole truth. Together, they show the category boundary. This is the real value of prompt audits. They reveal the words under which the market, or at least the answer environment, can recognize the company. They also reveal the words under which the company becomes someone else. If a brand only appears when the prompt uses its exact internal category language, the category is probably too fragile. If it appears for broad prompts but is misdescribed each time, the public source trail may be too vague. If it appears for software prompts even though the business sells a service, the website may be overusing platform language. If it disappears whenever the buyer adds a specific constraint, the company may lack public proof around that use case. A prompt audit is useful precisely because it is sensitive. The sensitivity becomes dangerous only when the team pretends it is stability. ## One answer is anecdote; repeated distortion is evidence A single bad answer is easy to overinterpret. The model may have retrieved an odd source. The prompt may have been worded poorly. The session may have carried context from earlier conversation. The tool may have decided not to search. The answer may change the next day. AI systems are variable enough that one screenshot should not become strategy. Repeated distortion is different. If several systems place the company in the wrong category, that is a signal. If different phrasings keep surfacing the same competitor set, that is a signal. If citation-based tools keep using the same third-party source, that source matters. If the company appears in branded prompts but not in category prompts, the market-facing evidence may be thin. If the AI answer gets the company’s current offer wrong in the same way a human buyer does, the problem is probably not only the model. The team should look less like a rank tracker and more like a field researcher. The question is not “what is our position?” The better question is “what kinds of questions cause the system to understand us correctly, and what kinds cause it to drift?” That drift is often where the work lives. ## A good prompt set has rough edges Clean prompt sets are suspicious. If every prompt is written in the company’s preferred language, the audit will flatter the brand. It will test whether AI systems understand the company when the buyer already talks like the company. Real buyers do not behave that way. They use old category names, competitor names, confused terms, procurement language, shorthand from analyst reports, and phrases picked up from colleagues. A useful prompt set includes some awkwardness. It asks the question the founder dislikes. It uses the competitor’s framing. It tests the term the sales team hears even though marketing avoids it. It includes “agency” if buyers keep saying agency. It includes “software” if the market keeps assuming software. It includes “alternatives to [competitor]” because buyers often know the competitor before they know the category. The work is not about creating a clean taxonomy. It is about approximating the messy ways people ask for help before they have learned the vendor’s language. This is also why prompt audits should include human interpretation. A spreadsheet can record whether a brand appeared. It cannot easily tell whether the answer would help a buyer make a decision, whether the description feels current, whether the competitor set is commercially meaningful, or whether the answer sounds confident because it has evidence or merely because the model writes confidently. Prompt data needs editorial judgment. Otherwise it becomes another dashboard pretending to know more than it does. ## The cited sources are often more important than the answer In citation-forward systems, the answer is only half the artifact. The sources are the other half. Perplexity describes itself as searching the internet in real time and summarizing information from sources, with citations that allow users to verify the answer. [Its help center](https://www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work) emphasizes source-backed responses. When a system like this recommends a competitor, the useful question is not only why the competitor appeared. It is which source made the competitor easier to recommend. Sometimes the answer cites a directory where the competitor has a complete profile and the company is absent. Sometimes it cites a category article that uses the competitor’s language. Sometimes it cites the competitor’s own page because that page is the clearest answer to the prompt. Sometimes the cited source is old, weak, or only partly relevant, but it is still the material the system found. A prompt audit that stops at mentions misses this layer. The citation path can show the surfaces that matter in a category: review platforms, industry lists, comparison pages, partner pages, analyst writeups, documentation, customer stories, or community discussions. These surfaces may be more actionable than the answer itself. The answer says what happened. The sources hint at why. ## Prompt audits should change public material, not just the prompt bank The weakest outcome of an AI visibility audit is a larger spreadsheet. A team tests prompts, records appearances, calculates a rough share of voice, and then repeats the exercise next month. The measurement becomes the work. The public materials remain the same. A useful audit should change the artifacts that answer systems and buyers encounter. If prompts expose category confusion, the homepage and service pages need clearer nouns. If competitor prompts reveal absent proof, the company may need case studies or third-party evidence. If citation paths keep pointing to directories, public profiles need attention. If the AI summary uses old language, the source trail needs cleanup. If the answer consistently misunderstands the service model, the site probably has the same problem for human readers. The prompt set is not the asset. The improved public explanation is the asset. There is no responsible way to guarantee that a particular AI system will mention a brand in a particular answer. The systems are too variable, and the mechanisms are too opaque. But a company can make the accurate version of itself easier to retrieve, easier to summarize, and easier to verify. A prompt audit is valuable when it points toward that work. ## The lens should be kept, but not worshipped AI visibility teams need prompt audits because otherwise they are guessing. The answer environment is now part of how buyers encounter brands, and companies should know what happens there. Ignoring it would be naive. The danger is turning the audit into a false map. A map implies a stable territory. A prompt audit is more like looking through a series of windows. Each window shows a part of the landscape, from a particular angle, in particular light. If several windows show the same broken fence, you should probably go outside and fix it. If one window shows a strange shadow, you should check before rebuilding the house. That is the discipline. Use prompts to notice. Use sources to investigate. Use patterns to prioritize. Then change the public material that made the pattern possible. The answer is not the market. It is a clue about how the market may be learning to ask. --- ## Why More Content Won’t Fix an Unclear Brand Source: https://zartrace.com/en/research/more-content-wont-fix-an-unclear-brand/ > Content can scale a clear idea. It cannot rescue a company that has not made itself understandable. A marketing team realizes the website is not producing enough qualified demand. The diagnosis arrives quickly: the company needs more content. A calendar appears. Keywords are grouped. Competitor blogs are reviewed. The team plans explainers, comparisons, trend pieces, thought leadership, glossary pages, and a few “ultimate guides” because someone still believes ultimate guides are allowed to exist without irony. Three months later, the site has more pages. Some of them are decent. A few rank. A few are shared internally with polite enthusiasm. The resource center looks less empty. The company is still hard to understand. A visitor can read two articles and still not know what the company actually sells. The blog explains the category better than the service page explains the offer. The thought leadership has a point of view, but the homepage does not turn that point of view into a buying path. The comparison post is useful, but the company never explains what it is comparing itself against. The content grows around the brand like ivy around a house whose front door is still hard to find. This is one of the quiet failures of B2B content strategy. The team produces material around the topic, but the company itself remains under-described. More content has not solved the clarity problem. It has distributed it. ## Content needs a center of gravity Every useful content program has a center. The company knows what it is. It knows who it serves. It knows what problem it solves. It knows which category it belongs to, even if that category is new. It knows what it is not. It knows what proof supports its claims. It knows how buyers compare it with alternatives. It knows which questions should be answered in public and which ones belong in a sales conversation. Without that center, content becomes inventory. The articles may be competent. The keywords may be reasonable. The topics may be adjacent to the business. But the reader cannot easily connect the material back to a clear vendor story. The company becomes visible around ideas without becoming understandable as a choice. This happens often in B2B because topic expertise can mask commercial ambiguity. A company may publish a sharp article about AI search, but if the service page does not explain whether it sells software, research, consulting, or managed operations, the article has only done half its job. It has earned attention without converting attention into comprehension. A content program should make the company easier to explain over time. If it does not, it is probably orbiting the business rather than clarifying it. ## Search-first content made the problem easier to ignore Search-informed content is not the villain. It can be useful, especially when buyer questions are real and the company has something specific to say. The problem begins when the search process becomes detached from the company’s actual clarity. Then the site fills with pages that answer adjacent questions but never sharpen the business itself. The team covers a topic cluster because competitors have covered it. It publishes definitions because the keyword tool says people search them. It writes comparison pages because the format seems commercially relevant. It produces a guide because the site “needs more authority.” This creates an illusion of progress. The library expands. Internal links multiply. The site looks more substantial. The buyer still cannot tell what the company does. Google’s guidance on helpful content is useful here because it cuts through a lot of content-production theatre. Google says its systems aim to prioritize helpful, reliable information created to benefit people, rather than content created primarily to manipulate rankings. [Google helpful content guidance](https://developers.google.com/search/docs/fundamentals/creating-helpful-content) That standard sounds obvious, but it is uncomfortable when applied honestly. Many B2B content programs are not created primarily to benefit people. They are created primarily to occupy topics. The reader is imagined, but not deeply respected. The article answers the surface query, avoids obvious mistakes, and says nothing that would surprise a competent buyer. The issue is not that the content is spam. The issue is that it is unnecessary. ## AI has made average content less defensible Generative AI has lowered the cost of producing acceptable prose. That changes the economics of content. A company can now generate twenty serviceable articles in a week. So can every competitor. So can every affiliate site, marketplace, directory, and solo consultant with a subscription and a prompt library. If the article has no original observation, no lived detail, no sharp example, no uncomfortable tradeoff, and no point of view, it enters a large gray pile of text that sounds plausible and matters little. Google’s guidance on AI-generated content does not ban AI use. It says generative AI can be useful for research and structure, while warning that using generative AI or similar tools to generate many pages without adding value for users may violate its scaled content abuse policy. [Google guidance on AI-generated content](https://developers.google.com/search/docs/fundamentals/using-gen-ai-content) The phrase “without adding value” is where the hard question lives. Value does not appear because a page is long. It does not appear because a keyword is present. It does not appear because a topic is relevant to the industry. It appears when the reader gets something they could not easily get from a generic summary: a clearer distinction, a better mental model, a concrete example, a useful argument, a credible measurement, a piece of evidence, a warning from experience. In AI-era content, average is closer to invisible than it used to be. The floor has risen. The ceiling has not. ## Unclear brands produce generic articles A company with unclear positioning tends to create generic content because the content team has no stable lens. If the business has not decided whether it is a tool, a service, a platform, a managed program, or a research layer, every article has to float above that uncertainty. If the company does not know which competitors it wants to be compared with, comparison content becomes either evasive or oddly broad. If the service pages do not explain what the buyer receives, educational articles cannot easily point toward a concrete next step. If proof is thin, the blog leans on commentary rather than evidence. The writers may be talented. The briefs may be thoughtful. The output still becomes soft because the brand underneath it is soft. This is why the hard work often has to happen before the calendar. The company needs to know which confusion it is trying to reduce. Maybe buyers do not understand the category. Maybe they compare the company with the wrong alternatives. Maybe the service model is unclear. Maybe the proof layer is weak. Maybe AI systems describe the brand using old language. Maybe the company’s external profiles contradict the website. Those are not article ideas yet. They are interpretation problems. Once the interpretation problem is named, content can help. A comparison essay can correct a bad market frame. A service explainer can make the buying path easier. A case study can turn a claim into evidence. A founder essay can define a category. A research piece can show the company’s method. A practical guide can earn trust because it solves a problem the reader actually has. The same title can be useful or useless depending on whether it is attached to a real confusion. ## Content should create public evidence, not just traffic A strong article does more than rank. It becomes public evidence of how the company thinks. After reading several good pieces from a company, a buyer should understand the business better. They should know what problem the team believes is underappreciated, what tradeoffs it sees, what it refuses to do, how it defines the category, where it disagrees with common advice, and why its approach is credible. That is different from content as inventory. Inventory fills a site. Evidence changes how the company is perceived. This matters for AI search too. The original Generative Engine Optimization research found that source visibility in generated answers could be improved by adding citations, relevant quotations, and statistics, though the effects varied by domain. [GEO research paper](https://arxiv.org/abs/2311.09735) The practical lesson is not to stuff articles with borrowed facts. It is that answer systems have more to work with when content contains substance: specific claims, support, examples, and clear relationships between ideas. An article that merely repeats the industry consensus gives both buyers and machines little reason to prefer it. An article that explains a problem with unusual clarity, names the hidden mechanism, and shows evidence becomes more useful as a source. The same is true for the brand. A company with a clear point of view creates content that compounds. Each piece makes the business easier to understand. A company without that clarity creates content that accumulates but does not compound. The library gets larger, but the brand does not get sharper. ## Sometimes the right move is fewer pages A company with an unclear brand may not need a bigger content plan. It may need five stronger public assets. A homepage that defines the category without hiding behind slogans. A service page that explains what the work includes. A proof page that shows what the company has actually done. A comparison piece that places the company correctly among alternatives. A serious editorial article that gives the company’s real point of view on the market. Those five pages can do more for discoverability than fifty thin articles around adjacent topics. This is difficult for teams because a calendar feels controllable. Clarity work is messier. It requires decisions. It may force the company to admit that a phrase everyone likes is not helping buyers. It may reveal that a service is still underdefined. It may require saying what the company does not do. It may require moving proof out of private decks and into public view. Content production can happen without those decisions. Good content strategy cannot. ## The question before publishing Before publishing an article, a company should ask whether the piece would still be worth reading if it never ranked. That question is not anti-SEO. A useful article can still target a real query. It can still bring organic traffic. It can still support a service page. It can still be structured well. But if the only reason the article exists is that a keyword tool found a gap, the article is probably weak. If it has no angle beyond “this topic is relevant,” it will sound like everything else. If it does not make the company easier to understand, it may be activity rather than strategy. A strong article earns its place in two ways. It helps the reader think better, and it helps the market understand the company more accurately. Anything less is just another page. --- ## How to Measure AI Visibility Without Turning It Into SEO Theater Source: https://zartrace.com/en/research/measure-ai-visibility-without-seo-theater/ > The danger is not that AI visibility cannot be measured. The danger is pretending it behaves like a clean ranking system when it does not. A marketing team receives its first AI visibility report. The deck looks serious. There is a share-of-voice chart, a mention-rate trend, a sentiment score, a table of prompts, and screenshots from ChatGPT, Perplexity, Gemini, and Google’s AI results. One slide says the company has improved from 12% to 19% visibility in thirty days. Another shows that a competitor has dropped. A third recommends content updates based on “prompt gaps.” The room feels relieved. The new channel has numbers now. Then someone asks what the numbers mean. Did visibility improve because the brand appeared in more answers, because the prompts changed, because the tool added a new source, because one model retrieved different pages, because the sample was small, or because a competitor’s name was counted differently? Does sentiment mean the answer was positive, or just that no criticism appeared? Is a citation better than a mention? Is a brand listed fourth in an AI answer equivalent to a search result in position four? What happens when the answer changes tomorrow? The relief thins out. This is where AI visibility measurement can either become useful or become theater. The useful version acknowledges instability and still extracts patterns. The theatrical version borrows the confidence of SEO dashboards without earning it. ## The old ranking metaphor breaks quickly Classic SEO measurement was never as clean as people pretended, but it had a familiar object: the search results page. Rankings varied by location, personalization, device, intent, and time, yet teams could still track positions, impressions, clicks, and pages with some shared understanding of what was being measured. AI answer environments are less tidy. The output may be a paragraph, a list, a comparison, a citation box, a carousel, a shopping unit, a set of links, or a blended answer whose structure changes by platform. One system may cite sources inline. Another may show links but not make them central. Another may produce an answer without visible citations. A fourth may change the response when the same prompt is asked in a slightly different way. OpenAI’s documentation says ChatGPT Search can provide timely answers with links to relevant web sources and may decide when to search based on what the user asks. [OpenAI Help](https://help.openai.com/en/articles/9237897-chatgpt-search) frames search as dynamic rather than a fixed results page. Google says AI Overviews and AI Mode may use query fan-out across subtopics and data sources, which means a single user question can produce a multi-source retrieval process. [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features) is explicit about this. Perplexity, meanwhile, presents itself as a source-backed answer engine where citations are part of the user experience. [Perplexity Help](https://www.perplexity.ai/help-center/en/articles/10352155-what-is-perplexity) describes answers as conversational and supported by verifiable sources. These are not different skins on the same SERP. They are different answer environments. That does not make measurement impossible. It means the measurement object has changed. ## Mentions are a start, not a diagnosis Most AI visibility tools begin with mentions because mentions are countable. Did the brand appear in the answer? Did the competitor appear? How often? In which prompt set? On which platform? That is a useful starting point. It is not enough. A brand can be mentioned in a way that hurts or confuses. It can appear under the wrong category. It can be listed as an alternative to a company it does not actually compete with. It can be named in passing while competitors receive fuller descriptions. It can appear in a broad prompt that attracts the wrong buyer. It can be cited as a source without being recommended as a vendor. The measurement has to preserve description quality. What did the answer say the company does? Was the statement accurate? Did it use old language? Did it include the right service? Did it mention proof? Did it cite the company’s own site, a third-party profile, a review platform, or an unrelated source? A mention without description is like seeing a name on a guest list without knowing whether the person was invited, tolerated, or confused with someone else. ## Tools are helpful, but their categories are still young The tooling market is moving quickly, which is both useful and dangerous. Adobe describes LLM Optimizer as a way to track brand mentions, citations, sentiment, and share of voice in AI-driven search, with dashboards for brand presence and competitive visibility. [Adobe Experience League](https://experienceleague.adobe.com/en/docs/llm-optimizer/using/essentials/overview) positions the product for marketing, SEO, and communications teams that need to understand how brands are presented in LLMs. Semrush’s AI Visibility Toolkit says it can benchmark brand visibility, analyze sentiment, discover prompts, track daily visibility, and identify competitive gaps. [Semrush documentation](https://www.semrush.com/kb/1493-ai-visibility-toolkit) frames this as an extension of search and competitive intelligence. Ahrefs describes Brand Radar as a way to monitor AI mentions and custom prompts across AI and search contexts, and its methodology discusses AI share of voice based on brand mentions and citations across major AI platforms. [Ahrefs Brand Radar](https://ahrefs.com/brand-radar) and its methodology material show how quickly the SEO tooling vocabulary is being rebuilt around AI answers. These tools are valuable because manual tracking becomes painful fast. A company cannot reliably manage hundreds of prompts, competitors, citations, and snapshots in a spreadsheet forever. Tooling helps create a repeatable observation layer. The risk is that dashboards make young measurements look more mature than they are. A visibility score can be useful if everyone understands its ingredients. It becomes theater if the score is treated as a natural property of the brand, like temperature, rather than as a constructed metric based on platform choices, prompt samples, weighting, and parsing rules. AI visibility metrics are not fake. They are just not self-explanatory. ## The prompt set is the measurement instrument In AI visibility, the prompt set is not a neutral container. It is the instrument. If the prompts are too branded, the company will look more visible than it is during discovery. If the prompts are too broad, the company may look weak in categories where it should not be competing. If the prompts use internal language, the audit may confirm the company’s private vocabulary rather than the market’s vocabulary. If the prompts exclude alternatives and trust questions, the measurement will miss important buying moments. This is why prompt design deserves more seriousness than it usually gets. A prompt set should represent buyer situations, not vendor hopes. It should include category discovery, alternatives, comparisons, use cases, trust checks, and named-brand questions. It should include the awkward language buyers use before they know the correct term. It should include the competitor names buyers actually mention, including the wrong ones. The prompt set should also be stable enough to track over time. If the team changes too much of the set every month, the trend line becomes a story about sampling, not visibility. New prompts can be added when the market changes, but the core set needs continuity. A bad prompt set can make a company look successful in a world buyers do not inhabit. ## Citations are not always better than mentions There is a natural instinct to treat citations as stronger than mentions. Often they are. A citation can show that a system used a page as source material. It can also create a pathway for the buyer to verify the answer. But citations need interpretation. A system may cite the company’s homepage while summarizing it badly. It may cite an old directory page because that page is easier to retrieve than a current service page. It may cite a competitor’s article that mentions your category but not your brand. It may cite a page that supports one sentence but not the larger conclusion. A citation can be a source of trust, a source of error, or a clue about what needs updating. Perplexity’s citation-forward interface makes source inspection especially natural, but the same habit matters anywhere sources are visible. Open the citation. Read what it actually says. Ask why the system might have used it. Ask whether the cited page gives the answer better material than your own site does. A citation count alone is too thin. Citation quality is where the useful work begins. ## Sentiment is fragile in B2B contexts Sentiment metrics can be helpful, but they are easy to overread. A brand described as “well-regarded” may receive a positive label even if the answer contains no evidence. A neutral summary may be commercially strong if it accurately names the category and cites the right service page. A mildly critical answer may be useful if it reflects a real limitation that buyers should understand. A competitor may receive positive language because the system is summarizing marketing copy, not because the market has judged the company favorably. B2B buying is full of cautious, specific judgments. “Good for mid-market teams with internal analytics support, but less suitable for companies needing fully managed execution” is more useful than a generic positive sentiment score. It tells the buyer something. It also tells the company where it is being placed. The measurement should care less about whether the answer sounds nice and more about whether it is accurate, specific, and aligned with the company’s actual fit. ## The best reports explain uncertainty without hiding behind it Bad reporting gives certainty where none exists. Equally bad reporting uses uncertainty as an excuse not to conclude anything. A useful AI visibility report sits between those failures. It says: this prompt set is limited but stable. These platforms behave differently. These answers may vary. These sources were visible in this run. These competitors appeared repeatedly. This description error showed up across multiple systems. This citation pattern suggests a surface gap. This trend is worth acting on; that one is too thin. The language should be plain. It should not pretend that AI answer tracking is a finished science. It also should not become so cautious that the client has no decision to make. The report should lead to edits: update these profiles, rewrite this service page, publish this proof, inspect these third-party sources, monitor this competitor, test these prompts again next month. Measurement that does not change the public record is just observation. ## Theatrical measurement has a recognizable smell SEO theater has always existed: dashboards with too many metrics, rankings without business context, traffic without quality, content scores without judgment, and reports that prove activity more than progress. AI visibility can recreate the same habits with newer vocabulary. The smell is familiar. A screenshot is treated as proof. A visibility score is shown without methodology. A prompt win is celebrated even though the prompt does not match the buyer. A sentiment label is treated as reputation. A citation is counted without being read. A month-over-month change is reported without acknowledging that the prompt set changed. A vendor claims “ChatGPT rankings” as if the answer environment were a stable SERP. The alternative is slower and more useful. Treat AI visibility as a set of observations about how the company is described, cited, compared, and supported across answer systems. Use tools where they help. Keep stable prompt sets. Inspect sources. Separate broad trends from anecdotes. Read the actual language. Ask whether the measurement reveals a public-record problem the company can fix. AI visibility can be measured. It just has to be measured with enough humility to remain useful. --- ## Interpretive Drift Is the New Brand Consistency Problem Source: https://zartrace.com/en/research/interpretive-drift-brand-consistency/ > A brand does not need every public sentence to match. It needs the public record to stop pulling the company into different meanings. A company finishes a rebrand in June. The new website is calmer, clearer, and more serious than the old one. The homepage describes the company as a managed AI visibility partner for B2B brands. The service pages explain audits, source-trail cleanup, and perception research. The founders are relieved. The old language finally feels gone. Then a buyer forwards a screenshot from an AI answer. The answer calls the company an “SEO engagement tool.” Nobody on the team has used that phrase in a year. A small investigation follows. The phrase does not come from the homepage. It comes from an old directory profile that described the company during its first positioning phase. The company’s LinkedIn page is current, but the Crunchbase-style startup profile still says “behavioral SEO signals.” A partner article describes the company as “user testing for search.” A podcast bio says “AI-powered SEO insights.” None of these descriptions is wildly wrong in isolation. Together, they create drift. That drift is now the brand consistency problem. Classic brand consistency worried about whether the logo, tone, tagline, and approved description matched across materials. That still matters. But in AI-era discovery, the harder problem is interpretive consistency: whether the public record leads buyers and systems toward the same basic understanding of the company. A brand can survive minor wording variation. It struggles when different surfaces imply different businesses. ## Consistency is not identical copy Some teams respond to drift by trying to make every public description identical. They paste the same paragraph into directories, social profiles, sales decks, founder bios, and partner pages. The result may be consistent, but it often feels dead. Real surfaces have different jobs. A LinkedIn summary should not read exactly like a directory description. A founder bio should not read exactly like an enterprise service page. Textual sameness is the wrong target. Semantic alignment is the useful one. A buyer should be able to move from the website to LinkedIn to a review profile to an AI summary and still feel that the company is the same kind of thing. The language can change. The level of detail can change. The emphasis can change. The underlying category, audience, offer, and proof should not quietly mutate. This distinction matters because rigid copy consistency is brittle. Teams stop maintaining it because it feels bureaucratic. Semantic consistency is more durable. It asks whether the surface preserves the meaning, not whether it repeats the sentence. A directory can say “AI visibility audits for B2B brands.” A service page can say “we examine how answer systems, search surfaces, directories, reviews, and public profiles describe your company.” A founder bio can say “I work on the human and source-trail side of AI discoverability.” These are not identical, but they point in the same direction. Interpretive drift begins when the directions diverge. ## Drift usually starts with reasonable compromises Most public inconsistencies are not the result of negligence at the beginning. They come from reasonable compromises that later become stale. An early-stage company uses “SEO” because buyers understand SEO. A directory requires a category, and the nearest available option is “marketing analytics.” A founder speaking on a podcast uses a casual shorthand because the audience is nontechnical. A partner writes a case study with their own framing. A sales deck uses bolder language than the website because it appears in a guided conversation. A local profile uses a shorter description because the field has character limits. At the time, none of this feels dangerous. Each surface is close enough. The danger appears when the company changes, or when AI systems and buyers begin assembling these surfaces together. A phrase that was harmless in one context becomes misleading in another. “SEO tool” may have been a useful shortcut in 2024. In 2026, after the company has become a managed AI visibility service, the same phrase pulls the brand into the wrong category. Drift is often a time problem. Old language remains attached to a current company. It is also a surface problem. Language written for one context travels into another. A podcast bio written for a friendly audience becomes a search result snippet. A directory category chosen quickly becomes a cited source. A partner’s shorthand becomes part of an AI answer. The web does not preserve the circumstances under which a phrase was reasonable. ## AI systems compress drift into confident summaries A human buyer might notice inconsistencies slowly. They see one description on LinkedIn, another on a review site, another on the homepage. The impression forms as a feeling: this company is hard to place. AI systems can compress that same inconsistency into a single paragraph. Google says AI Overviews and AI Mode may use query fan-out across subtopics and data sources to develop responses. [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features) describes this as a way to identify supporting web pages and display a wider set of helpful links than classic search. Perplexity says it searches the web and summarizes information from sources with citations. [Perplexity’s help center](https://www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work) frames the product around real-time search and source-backed answers. The practical implication is that public inconsistency can be assembled faster than before. A system may combine current homepage language, an old directory category, a third-party comparison, and a founder bio. The answer reads smoothly because the model writes smoothly. The underlying inputs may not be smooth at all. This is one reason AI-generated brand descriptions can feel uncanny. They are not always completely false. They are mosaics. One piece is current. One piece is old. One piece is adjacent. One piece is inferred. The result sounds authoritative because it has grammar, not because the brand is coherent. Interpretive drift is dangerous precisely because it produces plausible errors. ## Drift damages comparison before it damages conversion A buyer can tolerate a small inconsistency if the company is already shortlisted. They may ask a clarifying question. They may give the vendor a chance to explain. The bigger damage often happens earlier, when the buyer is still comparing options. If one source frames the company as software and another frames it as a service, the buyer may compare it with the wrong alternatives. If one profile suggests enterprise focus and another suggests small-business tooling, the buyer may misjudge fit. If a review platform preserves an old product category, the buyer may assume the current offer is less mature than it is. If AI answers describe the company with outdated language, the buyer may never click through to discover the correction. Drift therefore affects the comparison set. And the comparison set shapes everything. A company compared with the wrong alternatives may look expensive, vague, incomplete, or overbuilt. A managed service compared with a self-serve tool will look less scalable. A research-led program compared with a dashboard will look less measurable. A premium enterprise offer compared with lightweight software will look too complex. The company may still be good. It is being judged in the wrong room. This is why interpretive consistency matters commercially. It is not a matter of brand tidiness. It determines which mental shelf the buyer uses before evaluating value. ## The strongest source is not always the current one Teams often assume that the most recent description will dominate because it is most accurate. Public discovery does not guarantee that. An older source can be stronger if it ranks higher, is cited more often, appears on a trusted platform, uses clearer language, or sits inside a familiar taxonomy. A current website paragraph written in abstract brand language may lose to an old directory sentence written in plain nouns. A current service page may be correct but buried. A stale profile may be wrong but easy to extract. This is where consistency work becomes less like copy governance and more like weight management. The current version of the company has to become heavy enough to pull interpretation toward it. That may require updated owned pages, but also updated external profiles, recent third-party mentions, clearer service descriptions, refreshed founder bios, and public proof that supports the new positioning. If the only current source is the homepage, and the rest of the public web preserves older language, the homepage is carrying too much alone. Brand teams sometimes dislike this because the work feels small. Updating a directory does not feel strategic. Fixing a founder bio does not feel like a growth initiative. Rewriting a partner description does not feel like category design. But these small surfaces are where interpretive drift accumulates. AI systems are good at assembling small surfaces. Buyers are influenced by them too. ## How to think about consistency without becoming rigid A useful brand consistency system should preserve meaning while allowing surfaces to breathe. The company needs a canonical factual layer: what it is, who it serves, what problem it addresses, how it delivers value, what it does not claim, and what proof supports it. This layer should be boring enough to repeat. Around it, each surface can adapt. A homepage can be more expansive. A directory profile can be concise. A founder bio can be personal. A comparison article can be analytical. A service page can be operational. The question is whether all of them lead the reader toward the same company. This is a more realistic standard than identical copy. It also creates better material for AI search because different surfaces can contribute compatible pieces of the same story. The hard part is maintenance. Someone has to know where the brand lives publicly. Someone has to revisit those surfaces after a positioning change. Someone has to notice when a third-party source keeps an old category alive. Someone has to decide whether a description is merely different or meaningfully misleading. That work is not glamorous. It is infrastructure. ## Drift will always return No company can eliminate interpretive drift permanently. Public surfaces decay. Teams change language. Competitors create new frames. Platforms add categories. AI systems retrieve different sources. Buyers use unexpected terms. A perfectly controlled brand environment was never realistic. The practical aim is to keep drift from becoming the dominant interpretation. A healthy public brand record has some variation. It can absorb different contexts without losing the company’s shape. The website, profiles, reviews, comparison pages, and AI summaries do not need to sound identical. They need to keep pointing toward the same commercial reality. When they stop doing that, the brand has not merely become inconsistent. It has started to mean different things depending on where the buyer meets it. --- ## The Hidden Places Buyers Check Before They Contact You Source: https://zartrace.com/en/research/hidden-places-buyers-check-before-contacting-you/ > By the time a buyer appears in your CRM, they may already have met your company in five other places — some of which your team has not looked at in months. A buyer fills out a contact form at 4:37 p.m. on a Thursday. In the CRM, it looks like a new lead. The source field says organic search. The page path shows a visit to the homepage, then the pricing page, then the contact page. It seems straightforward enough. The actual path was messier. Two days earlier, the buyer asked an AI assistant for vendors in the category. One competitor appeared with a short description and three citations. Your company appeared too, but the answer called it a “consulting firm,” which was not quite right. The buyer searched your brand name, opened LinkedIn, checked whether the founder looked real, skimmed a review profile that had not been updated since an old product launch, found a partner page with a better explanation than your homepage, and forwarded a comparison article to a colleague. The colleague replied with a screenshot from another AI answer that omitted you entirely. By the time the form was submitted, the buyer was not starting a conversation. They were continuing one you had not been part of. This is the modern B2B discovery problem. The website still matters, but it is no longer the only room where evaluation happens. Buyers walk through a building made of search results, AI answers, directories, review platforms, social profiles, old articles, partner pages, community discussions, and whatever else the public web offers them. Some of those rooms are well maintained. Some have old furniture and bad lighting. ## Buyers want witnesses, not only the official story A company’s website is the official version of the brand. It is the place where positioning becomes clean, where the offer is framed carefully, where proof is selected, and where inconvenient complexity is edited into a coherent narrative. Buyers know this. They expect the website to be optimistic. That does not make the website useless. It makes it insufficient. A serious buyer usually wants witnesses. A review platform is a witness. A directory is a witness. A partner page is a witness. A founder interview is a witness. A customer quote is a witness. A LinkedIn profile is a witness. An AI answer is becoming a kind of witness too, although an imperfect one. These witnesses are not equally reliable. Some are thin. Some are stale. Some are biased. Some are just wrong. Buyers use them anyway because triangulation feels safer than trusting the vendor alone. NN/g’s work on web trustworthiness includes “connection to the rest of the web” as one of the factors that helps a website communicate credibility, alongside design quality, up-front disclosure, and comprehensive current content. [NN/g on trustworthy design](https://www.nngroup.com/articles/trustworthy-design/) That idea has aged well. A company that exists only inside its own marketing copy feels lighter than a company whose story is supported elsewhere. The second company may simply be easier to verify, even when its actual product is no better. ## The hidden surfaces are usually not hidden from buyers Teams often call these places “hidden” because they are hidden from internal process, not from the market. Nobody owns the old directory profile. The LinkedIn page was updated during the rebrand, but the founder bios were not. The review platform belongs to the customer marketing team, except that team only checks it before renewal season. The partner page was written by another company. The marketplace profile was created during a pilot. The comparison article is old enough that nobody wants to spend energy correcting it. The community thread is not flattering, but it is not bad enough to become a priority. From the buyer’s perspective, all of these are simply the public web. They do not care which team owns the surface. They care whether the company looks coherent. A buyer may not consciously register every inconsistency. They may not write down that the directory says “SEO software,” the homepage says “AI visibility infrastructure,” and the review site says “marketing analytics.” They just feel that the company is harder to place. The category gets fuzzy. The offer seems less mature. The next conversation feels riskier. This is how trust leaks: through small public mismatches that make the buyer do more interpretive work. ## AI answers have become part of the evidence path AI search adds a new surface to the buyer’s investigation. It is not simply another search box. It is a summarizer of other surfaces. A buyer can ask what your company does, whether it is credible, how it compares with alternatives, which vendors belong in the category, or what to consider before choosing a provider. The answer may include citations, or it may not. Either way, the buyer receives an interpretation before reaching your site. This matters because AI-assisted buying is no longer marginal in some B2B categories. Gartner reported in March 2026 that 67% of B2B buyers prefer a rep-free experience, and 45% of surveyed buyers had used AI during a recent purchase. G2’s April 2026 software-buyer research reported that 51% of B2B software buyers now begin software research with an AI chatbot more often than with Google, up from 29% in April 2025. [Gartner source](https://www.gartner.com/en/newsroom/press-releases/2026-03-09-gartner-sales-survey-finds-67-percent-of-b2b-buyers-prefer-a-rep-free-experience) [G2 source](https://www.prnewswire.com/news-releases/new-g2-research-half-of-b2b-software-buyers-now-start-their-research-with-ai-chatbots-302742807.html) Survey numbers should be treated with some care. They vary by category, geography, sample, and wording. Still, the pattern is obvious enough: buyers are increasingly asking systems to synthesize vendor information before they speak with vendors. This changes the role of every public surface. A stale directory profile is no longer only a stale directory profile. It may become raw material for an answer. A partner page is no longer only a partner page. It may become a citation. An old review profile is no longer only a place someone might check after shortlisting you. It may influence whether you make the shortlist at all. The public web has become a pre-sales room, and AI systems are rearranging the furniture. ## The website cannot compensate for every neglected room It is tempting to solve this by making the website better. That is often necessary. A clear homepage, strong service pages, current proof, and useful comparison content all matter. But the website cannot fully compensate for an incoherent public trail. If the homepage says one thing and the rest of the web says three other things, buyers may treat the homepage as the polished exception. If the website claims enterprise credibility while external surfaces show a thin or outdated presence, the claim feels less stable. If the site describes a new category but third-party sources keep pulling the company into an old one, the buyer has to decide whom to believe. This is where many teams underestimate the informational life of external pages. A directory profile may be ugly and generic, but it sits in a taxonomy buyers recognize. A review site may have imperfect data, but it feels less vendor-controlled. A partner page may be short, but it confirms a relationship. A third-party article may be outdated, but it carries the authority of being outside the company’s own domain. The homepage is the official statement. The other surfaces are the receipts, however incomplete. ## Some surfaces matter because competitors are there A hidden surface becomes especially important when competitors are represented better than you are. Imagine a buyer asking an AI assistant for providers in a category. The answer cites a review platform, an industry list, and a comparison article. Three competitors appear across those sources. Your company is absent from two of them and miscategorized in the third. The buyer does not need to know the mechanics. The effect is simple: competitors look more established. The problem may not be content quality in the narrow sense. It may be surface coverage. Some categories have obvious surfaces: G2, Capterra, Clutch, Gartner Peer Insights, Google Business Profile, industry associations, local directories, marketplaces, partner ecosystems. Others are messier: analyst blogs, Reddit threads, LinkedIn posts, niche newsletters, GitHub pages, YouTube reviews, founder interviews, vendor comparison pages. Every category has its own map of places where buyers look for confirmation. Most companies know the obvious surfaces. Fewer know which ones AI systems and serious buyers actually use when forming a category-level answer. That is why surface work should begin with observation, not assumption. Look at what appears when buyers ask category questions. Look at what competitors are cited for. Look at which third-party pages rank for “[category] alternatives,” “[competitor] reviews,” and “[brand] vs.” The important surfaces often reveal themselves by being used. ## The neglected surfaces are also where old versions survive External surfaces preserve old versions of companies because nobody has an incentive to keep them perfectly current. A review platform may preserve an old product category because changing it requires platform approval. A partner page may remain untouched because the partner has bigger priorities. A podcast bio may live forever because nobody edits old show notes. A local listing may have the wrong description because it was set up by a former employee. A startup database may reflect the fundraising story, not the current business. A comparison article may keep the old feature list because updating it is not worth the publisher’s time. None of these sources is malicious. Most are just unattended. The problem is that unattended pages still speak. They speak to buyers who are trying to verify the company. They speak to search engines. They may speak to AI systems when those systems retrieve or summarize public sources. A company can change internally and still have its old commercial shape preserved in dozens of small public places. This is why brand maintenance increasingly resembles archive maintenance. The team has to know what it wants to say now, and also what the web is still capable of saying on its behalf. ## A serious buyer reads the gaps Buyers rarely evaluate public evidence with perfect rationality. They do not run formal audits of your source trail. They skim, infer, and move on. But they read gaps. A company says it serves enterprise clients, yet there are no enterprise examples. It says the product is mature, but the public profiles look early-stage. It says the service is transparent, but there is no sample output. It says it works globally, but external listings show one market. It says it is different from an agency, but every third-party source calls it an agency. Each gap creates a small private question. The buyer may bring the question to sales. More often, the question becomes part of a general feeling: this vendor is harder to verify. The company that wins is not always the one with the most public material. It is often the one with fewer unexplained gaps. The site says what the company does. The external profiles confirm it. The reviews do not contradict it. The AI summary gets the basics right. The directory category makes sense. The founder profile supports the positioning. The case study shows the work. The pricing or scope context reduces uncertainty. The comparison content helps the buyer place the vendor correctly. This is not perfection. It is coherence. ## The path should be maintained before the lead appears The mistake is waiting until a buyer asks. By the time the buyer asks why an old profile says something different, the damage has already happened. The company has become harder to trust before the conversation begins. The sales team now has to spend time correcting the public record instead of advancing the deal. A better habit is to audit the path before the buyer walks it. Start with the branded search result. Then the AI summary. Then LinkedIn. Then review platforms. Then directories. Then category pages. Then competitor comparisons. Then partner mentions. Then old articles. Then the places AI tools cite when they answer category questions. Read them as a cautious buyer, not as a marketer defending the brand. Total control is impossible. The useful work is knowing which surfaces matter, which ones are stale, which ones contradict the website, and which ones create unnecessary doubt. A buyer who contacts you has already been somewhere. The question is whether those places helped you. --- ## The Comparison Page Is Becoming a Trust Document Source: https://zartrace.com/en/research/comparison-page-trust-document/ > A serious comparison page does not win by flattering the company. It wins by making the buyer feel less manipulated. A buyer searches for two vendors late in the buying process. The query is simple: one company name, “vs,” another company name. The buyer has probably seen a demo, read a few pages, or heard both names from colleagues. They are not looking for a generic category introduction anymore. They are trying to make a decision without inviting a salesperson into the room too early. The first result is a vendor-owned comparison page. It opens with a confident headline explaining why the vendor is the better choice. The competitor is described just enough to be dismissed. Every row in the comparison table favors the page owner. The page includes a few vague feature checks, a testimonial, and a call to book a demo. It is not false exactly. It is simply too clean. The buyer can feel the choreography. So they leave and look for another source. This is the failure mode of many comparison pages. They were created as SEO assets, but buyers read them as trust documents. The page may rank. It may capture a commercially valuable query. It may even produce demos. Still, it leaves a residue of suspicion because it pretends the decision is easier than it is. In AI-assisted discovery, that problem becomes sharper. Comparison pages are no longer just landing pages for users who click. They can become source material for AI answers that summarize differences between vendors, categories, and approaches. If the page is thin, one-sided, or evasive, it gives both humans and systems poor material. ## Buyers do not come to comparison pages innocent A person reading a comparison page usually knows it has an agenda. That does not make the page useless. Buyers can tolerate bias when the bias is visible and the information is still useful. What they dislike is false neutrality. A vendor can say, in effect, “we have a point of view, and here is how we see the tradeoff.” That feels more credible than pretending to conduct an independent evaluation in which every conclusion conveniently favors the author. The buyer’s real question is rarely “which vendor is universally better?” It is more specific and more practical: better for whom, under what constraints, at what stage, with what internal resources, and at what cost of switching? A good comparison page respects that question. It gives the reader a way to understand the shape of the decision. It explains where the competitor may be strong. It names the conditions under which the company itself is a better fit. It makes the boundary legible. A weak comparison page treats the buyer as if they are looking for permission to choose the page owner. That difference is easy to sense. It is also measurable in the quality of the prose. Pages that never concede anything tend to sound like advertising. Pages that can describe a competitor fairly often sound like they were written by someone who understands the market. ## The old comparison table is not enough The comparison table became popular because it is easy to scan. It also became lazy. A row says “AI visibility monitoring.” One vendor has a checkmark. The other has a dash. Another row says “human research.” Another checkmark. Another dash. The pattern continues until the page owner has won every meaningful category. Some tables are useful. Many are decorative. The problem is not the table itself; it is the lack of explanation around it. A buyer does not only need to know whether a feature exists. They need to know whether that feature matters, how it is implemented, what tradeoff it introduces, and whether it changes the buying decision. A comparison page that only lists features is often avoiding the harder conversation. If Vendor A is better for teams that need self-serve dashboards and Vendor B is better for teams that need managed interpretation, say that. If one product is cheaper but requires more internal effort, say that. If one service is stronger for enterprise governance and weaker for speed, say that. These are the details buyers actually use. Nielsen Norman Group’s concept of “information scent” is useful here. Users choose links and paths based on cues that suggest whether the destination will answer their question. [NN/g describes information scent](https://www.nngroup.com/articles/information-scent/) as a mix of link labels, context, and prior experience. A comparison page with strong scent gives the buyer confidence that the page understands the decision they are trying to make. A page with weak scent gives them a list of vendor-approved attributes and asks them to do the real comparison themselves. ## Fairness is a conversion asset Companies sometimes fear that fairness will weaken a comparison page. If they admit a competitor is better in some cases, will buyers leave? Some will. That may be the point. A comparison page should not try to capture every buyer. It should help the right buyer understand the tradeoff sooner. If the competitor is genuinely better for a certain use case, saying so can make the page more credible for the use cases where your company is stronger. The buyer sees that the author is capable of making distinctions rather than simply selling. This is especially important for complex B2B categories. A vendor may be better for large enterprise teams but worse for early-stage companies. A managed service may be better when the client lacks internal capacity but worse when the client needs direct software control. A research-led approach may be better for interpretation problems but unnecessary for a company that only wants raw tracking metrics. A page that explains those distinctions can still be persuasive. It persuades by clarifying, not by cornering. Trustworthiness research from Nielsen Norman Group has long emphasized up-front disclosure, current content, and connection to the rest of the web as credibility factors. [NN/g’s discussion of trustworthy design](https://www.nngroup.com/articles/trustworthy-design/) is not about comparison pages specifically, but the logic applies cleanly. A page that discloses real differences, stays current, links outward where appropriate, and provides enough evidence feels less like a trap. In a market saturated with vendor content, restraint becomes a trust signal. ## AI systems need comparison pages that do not lie by omission AI search introduces a second reader for comparison content. When a buyer asks an AI tool to compare two vendors, the system may draw from company pages, third-party articles, directories, reviews, and other available sources. Google says AI Overviews and AI Mode may use query fan-out across subtopics and data sources to develop a response. [Google’s documentation](https://developers.google.com/search/docs/appearance/ai-features) describes this as a way to surface wider and more diverse links than classic search. Perplexity describes its answers as source-backed summaries, which means the public comparison material around a category can become part of the answer environment. A one-sided comparison page can still be used, but it may be less useful than a page that clearly states dimensions of difference. Systems need extractable distinctions. “We are better” is a poor distinction. “This approach is better when the team needs managed interpretation; the alternative is better when the team needs self-serve monitoring” is more usable. The second sentence also helps a human buyer. This is where AI readability and human honesty overlap. The page should not become robotic or artificially neutral. It should make the comparison legible enough that a person or system can summarize it without flattening the decision into nonsense. A comparison page that lies by omission creates bad source material. It may omit the competitor’s strengths, omit its own weaknesses, omit price differences, omit implementation requirements, or omit the fact that the two vendors solve slightly different problems. The answer system may then produce a shallow comparison, and the buyer may notice that something feels missing. ## The best comparison pages show the buying situation A useful comparison page often begins before the vendors. It explains the buying situation. Why would someone compare these two companies in the first place? Are they both in the same category, or only adjacent? Is the buyer choosing between software and service, enterprise and mid-market, automation and interpretation, speed and governance, depth and cost? This framing matters because many comparisons are category mistakes. A buyer may compare an AI visibility audit service with an SEO rank-tracking tool. They may compare a managed user research program with a traffic-generation vendor. They may compare a digital PR agency with an answer-engine optimization platform. The comparison is not stupid. It reflects the buyer’s uncertainty. A good page does not mock that uncertainty. It uses it. The page can say, in plain language, why the comparison happens and what the buyer should pay attention to. It can explain that both options relate to visibility, but one measures mentions while the other investigates interpretation. Or that both relate to trust, but one handles reputation repair while the other builds public evidence before a reputation problem appears. This kind of explanation does more than sell. It teaches the category. ## Evidence should sit near the claim Comparison pages often make claims without enough proof. They say one vendor is more transparent, but show no reporting example. They say implementation is faster, but do not explain what implementation includes. They say support is better, but offer no service model. They say the competitor is less flexible, but give no meaningful context. The buyer is left with adjectives. A more serious page places evidence near the contested claim. If the difference is reporting, show what gets reported. If the difference is methodology, describe the method. If the difference is source coverage, explain which surfaces are checked. If the difference is human research, describe what participants do and what the output looks like. This does not require revealing proprietary detail. It requires enough specificity that the buyer can evaluate the claim. Google’s helpful content guidance asks whether content provides original information, reporting, research, or analysis, and whether it offers substantial value compared with other pages in search results. [Google Search Central](https://developers.google.com/search/docs/fundamentals/creating-helpful-content) frames helpful content around people-first value rather than search manipulation. A comparison page that merely repeats a vendor’s sales claims is unlikely to meet that standard in any meaningful editorial sense. A comparison page has to earn its existence because it is inherently suspect. ## The page should age like a maintained document Comparison pages decay quickly. Competitors change pricing, launch features, sunset services, move upmarket, acquire companies, rewrite positioning, or enter new regions. A comparison page from last year can become quietly misleading even if it was written carefully at the time. This decay is dangerous because comparison pages are often high-intent pages. They appear late in the buyer journey, when the reader is closer to a decision and more sensitive to trust. An outdated claim on a generic blog post is annoying. An outdated claim on a comparison page feels manipulative. The page should therefore behave more like a maintained document than a campaign asset. It should have a visible update rhythm, careful language around changeable facts, and enough humility to avoid claims that will rot quickly. “As of April 2026” is often more credible than a timeless assertion when the market is moving fast. In AI search, maintenance matters for another reason. Old comparison pages can become part of the source trail. If they are wrong, they do not simply mislead visitors who click. They may also contribute to summaries, citations, and competitor descriptions elsewhere. A neglected comparison page can keep arguing long after the company has stopped believing the argument. ## The comparison page is a test of how honest the category is A company that writes a good comparison page has to understand the market well enough to be fair. It has to know its own edges. It has to know where competitors are strong. It has to know which differences matter to buyers and which are cosmetic. That is why comparison pages are revealing. They show whether the company can think beyond its own pitch. A thin page says, “choose us.” A better page says, “here is the decision you are really making.” The second version may still lead to a sale. More importantly, it gives the buyer a reason to trust the company before the sales conversation begins. --- ## How to Check What AI Systems Say About Your Brand Without Fooling Yourself Source: https://zartrace.com/en/research/check-what-ai-systems-say-about-your-brand/ > A brand audit in AI search starts to become useful only after you stop treating one answer as a verdict. A VP of marketing opens ChatGPT after a board meeting and types the company name. The answer is not terrible. It gets the category roughly right, mentions two real products, and links to the website. It also describes the company as “early-stage,” although the business now has 140 employees and enterprise customers. The VP takes a screenshot, drops it into Slack, and asks whether this is “an AI visibility issue.” Within ten minutes, three more screenshots appear. Someone asks Perplexity the same question and gets a cleaner answer with citations. Someone asks Google’s AI Mode for alternatives to a competitor and the company is missing entirely. Someone asks Gemini a broader category question and sees two smaller competitors listed above them. The thread becomes animated in the way internal Slack threads become animated when nobody is sure whether the evidence is serious. There is a small embarrassment inside this scene. Everyone knows the screenshots matter, but nobody knows how much. That is where most AI brand checks begin: a few prompts, a few screenshots, a little panic, and a large temptation to overinterpret. The answer may be wrong, but it may also be unstable. The brand may be absent, but perhaps the prompt was poorly chosen. A competitor may appear, but perhaps the question was closer to their segment than yours. A citation may look authoritative while pointing to an old directory page nobody has updated in years. Checking what AI systems say about your brand is worth doing. It is becoming part of ordinary brand hygiene. But done casually, it produces more anxiety than knowledge. ## The first answer is usually the least interesting one Most teams start with a branded prompt: “What is [Company]?” That is a natural place to start, but it is also a narrow one. A branded prompt tests whether the system can describe a company after being directly named. It does not tell you much about whether the company appears during discovery. Buyers rarely begin with the exact question marketing teams wish they would ask. They ask for a vendor category. They ask for alternatives. They ask whether a company is trustworthy. They ask how one provider compares with another. They ask which solution fits a specific problem, budget, market, or team size. They ask a question that sounds slightly wrong because they do not yet know the category language. The branded answer is still useful. It shows the basic public biography of the company: what the system thinks the company does, which products or services it names, whether the facts are current, and whether it uses language the company recognizes. But the more revealing answers often come from prompts where the brand is not named at all. A company may be described accurately when named and still be absent when the buyer asks for providers in the category. Another may appear in “alternatives to [competitor]” prompts but not in “best providers for [use case]” prompts. A third may appear only when the prompt uses old category language. These patterns tell different stories. The useful audit begins when the team stops asking, “What does AI say about us?” and starts asking, “In which buyer situations does AI remember us, and in which ones does it not?” ## Different systems expose different parts of the public record It is also a mistake to treat “AI” as one surface. ChatGPT Search can provide timely answers with links to relevant web sources, and OpenAI says ChatGPT may decide to search the web depending on the question or the user’s choice. [OpenAI’s help page](https://help.openai.com/en/articles/9237897-chatgpt-search) is careful to frame search as something that can bring web sources into the answer environment, not as a guarantee that every answer is produced in the same way. Perplexity describes itself as an AI-powered search engine that searches the web and returns conversational answers backed by verifiable sources and citations. [Its help center](https://www.perplexity.ai/help-center/en/articles/10352155-what-is-perplexity) makes citations central to the product experience. Google’s AI features may use query fan-out, issuing multiple related searches across subtopics and data sources to build a response, according to [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features). Those mechanisms are not identical. They do not retrieve the same sources, present the same citations, or respond to prompts with the same level of stability. Even within one system, answers can change by phrasing, freshness, location, session context, and whether retrieval is triggered. This is irritating if you want a single score. It is useful if you want diagnosis. Perplexity may expose which public pages are being treated as sources because citations are part of the interface. Google’s AI features may reveal how a broad comparison query gets decomposed into related subtopics. ChatGPT may show where a named brand is understood well but where category-level discovery is weak. A system that refuses to mention the company may not be “wrong” in isolation; it may be showing that the brand’s public evidence is too thin for that particular question. No single platform deserves to be crowned as the truth. The useful work is reading across platforms for recurring patterns. If one system misses the company, it is a clue. If every system misses it for the same buyer question, it is a problem. ## A good prompt set sounds like a buyer, not a vendor A common failure mode in AI visibility checks is vendor-centered prompting. The team writes prompts using internal language: exact product names, preferred category labels, polished phrasing from the homepage. The system responds reasonably well, and everyone feels better than they should. A real buyer is usually less precise. A procurement lead may ask for “tools that monitor what ChatGPT says about a brand” even if the company calls the category “AI visibility intelligence.” A founder may ask for “how to know if AI search is hurting our company” rather than “generative engine optimization audit.” A marketer may ask for “why Perplexity recommends competitors” or “how to track brand mentions in AI answers.” A board member may ask for “companies helping B2B brands show up in AI search,” which is broad enough to pull in SEO agencies, analytics tools, reputation firms, and newer GEO providers. Those imperfect prompts are valuable because they resemble the market’s language before the market has adopted yours. The audit should include branded questions, but also category questions, alternative questions, comparison questions, trust questions, and use-case questions. It should include prompts that use your preferred language and prompts that use the messy language buyers actually use. It should include direct competitors and the wrong competitors you keep getting compared with, because AI systems may reproduce the same confusion. The uncomfortable prompts are often the best ones. They show whether the company is discoverable only inside its own vocabulary. ## The source is often more important than the answer Screenshots are seductive because they make a variable answer look fixed. A screenshot has edges. It feels like evidence. But the more useful object is usually the source behind the answer. When an AI system cites a directory, a review platform, an industry article, a partner page, or a service page, open it. Read it like a buyer who is trying to decide whether the company belongs in the category. Does the page describe the current business? Does it use old language? Does it include competitors but not you? Does it frame the category in a way that favors one type of vendor? Does it use a concrete description your own site avoids? Often the AI answer is less mysterious after the cited pages are inspected. A competitor is recommended because it appears in three third-party sources for the category. Your company is missing because it is absent from the source the answer used. The answer gets your offer wrong because the only page with a plain description is outdated. The AI cites your homepage but still summarizes the company vaguely because the homepage itself never says what the buyer receives. A citation is not a trophy. It is a doorway into the source trail. This is where an AI visibility check becomes more than prompt play. It tells the company which public materials are shaping its description and which materials are failing to do so. ## Do not measure only presence The simplest AI visibility metric is whether the brand appears. It is also the easiest metric to misuse. Presence can be good, bad, or irrelevant. A brand appearing in the wrong category may be worse than absence. A brand mentioned in a neutral list without proof may not influence a buyer. A brand cited as a source may matter more than a brand merely named in passing. A competitor appearing above you may matter in one prompt and not matter in another if the prompt is poorly matched to your actual buyer. The audit should read the answer as a piece of market evidence. How is the brand described? Is the category right? Are the facts current? Which competitors appear? What proof is named? Are sources shown? Does the answer sound confident, cautious, or confused? Does the system repeat old language? Does it omit the company’s strongest use case? A company can have high mention frequency and low description quality. Another can have low mention frequency but strong citations when it does appear. Another may be invisible for broad category prompts but strong for specific use cases. These are different conditions, and they require different work. The dashboard temptation is to compress them into one score. The editorial task is to keep enough texture to know what the score means. ## Stable prompts beat dramatic screenshots The most useful AI visibility checks are boring in one specific way: they repeat. A stable prompt set, run on a schedule, reveals movement. A random set of screenshots reveals mood. If the team changes every prompt every month, it cannot tell whether the brand changed, the market changed, or the prompt changed. If the team keeps the core prompts stable, patterns begin to show. The prompt set will still evolve. New services launch, competitors appear, buyers adopt new language, and platforms change. The core set, however, should remain steady enough to compare one month with the next. It also helps to preserve the full answer, not only the conclusion. AI answers are variable and sometimes slippery. A result that looks like a win in a screenshot may contain a factual error in the second paragraph. A result that looks like a loss may cite a source that reveals exactly where to improve coverage. The archive of answers becomes useful only when the team can inspect it later. The best rhythm is modest. Run the prompts. Read the answers. Inspect the sources. Note the patterns. Decide which public materials need to change. Then run the same prompts again after enough time has passed for the public record to shift. This is slower than screenshot theater. It is also the only version that compounds. ## The audit should end with edits, not awe An AI visibility check is not complete when the team knows what the systems said. It is complete when the team knows what to change. Sometimes the change is factual. A profile needs updating. An old page needs a redirect. A directory category is wrong. A service name is outdated. Sometimes the change is explanatory. The service page needs to describe the deliverable. The homepage needs a plainer first sentence. The comparison article needs to explain why the company is not in the category buyers assume. Sometimes the change is evidential. The company needs proof in public, not only in sales decks. It needs a case study, a sample output, a more current partner page, a review profile that reflects the current business, or third-party material that supports the category. And sometimes the finding is that the prompt itself is not yours to win. That is useful too. AI visibility should not become a project to appear everywhere. It should help the company appear where it belongs and understand where it is being misplaced. The practical value of an AI brand check is not the screenshot. It is the correction to the public record that the screenshot made visible. --- ## A Buyer’s Question Is Not a Keyword Source: https://zartrace.com/en/research/buyers-question-is-not-a-keyword/ > Keywords describe what people type. Buyer questions reveal what they are trying to resolve. A marketing team opens the same spreadsheet it has used for years. The columns are familiar: keyword, volume, difficulty, intent, current rank, target URL. The terms are not wrong. They are the phrases the category has always cared about. Some are commercial, some informational, some belong to competitors. The team uses them to brief new pages and update old ones. Then someone asks ChatGPT a question that no keyword list would have produced: “We are a 120-person B2B SaaS company replacing a patchwork of user research tools and customer interviews. Which vendors should we look at if we care more about evidence quality than dashboards?” The answer does not behave like a search result. It does not simply match a phrase. It reconstructs a situation. It names a company the team expected, misses another one it should have included, pulls in a research agency that only partly fits, and cites a page from a competitor’s site that was never designed as a primary landing page. One of the citations is a comparison article. Another is a product page. A third is an old guide that happens to explain the evaluation problem clearly. The team looks back at the spreadsheet. The keywords are still useful, but they suddenly feel too flat. This is a recurrent pattern in AI-era discovery. Companies keep optimizing around nouns while buyers increasingly ask in situations. A keyword may name the category. A buyer question describes a problem under constraints: company size, risk tolerance, existing stack, internal politics, budget ambiguity, trust requirements, and the thing the buyer does not yet know how to name. ## The old keyword gave the page a target. The new question gives it a scene. Classic search did not ignore context, but the interface encouraged compression. A buyer who needed help choosing a vendor often reduced the problem to a phrase: “best customer research tools,” “AI visibility audit,” “B2B website clarity,” “brand monitoring software,” “GEO agency.” The search engine returned pages, and the buyer did the work of stitching them into a decision. AI search changes the shape of the exchange. The buyer can bring more of the messy situation into the query. They can ask for tradeoffs, constraints, comparisons, and exceptions. They can say they are not ready for enterprise software, or that they need something a marketing team can operate without a data scientist, or that they already tried a cheaper tool and the internal team stopped trusting the outputs. The platform documentation points in the same direction. [OpenAI’s ChatGPT Search help page](https://help.openai.com/en/articles/9237897-chatgpt-search) describes timely answers with links to relevant web sources and says ChatGPT may choose to search depending on the question. [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features) says AI Overviews and AI Mode may use query fan-out across related searches and sources. [Perplexity’s help center](https://www.perplexity.ai/help-center/en/articles/10352155-what-is-perplexity) describes conversational answers backed by verifiable sources and citations. Those platform details matter because they point to a broader editorial shift. The unit of discovery is no longer only the typed phrase. It is the question environment around the phrase. A buyer asking “best AI visibility tools” may want a software comparison. A buyer asking “how do I know what ChatGPT says about our brand before we hire a GEO agency?” may need an audit method, not a vendor shortlist. A buyer asking “why does Perplexity cite our competitor but not us?” is somewhere else entirely. All three may touch the same commercial category. They should not produce the same page. ## A keyword can hide the disagreement inside the buying group B2B search terms often look cleaner than the internal conversation that created them. Inside a buying group, one person may think the problem is SEO. Another thinks it is reputation. A third wants better content. A fourth worries that AI tools are misdescribing the company. The executive sponsor does not care which label wins; she wants fewer confusing vendor conversations and more confidence that the company appears correctly in public discovery. By the time someone opens a browser, the internal disagreement may have been compressed into a short phrase like “AI search optimization.” A page optimized only for that phrase can miss the real tension. It may define the category, compare tools, list tactics, and mention ChatGPT. It may still fail because the buyer’s actual question is not “what is AI search optimization?” but “which part of our visibility problem is caused by the website, which part by external sources, and which part by AI systems repeating old information?” That question has a different shape. It needs diagnosis before prescription. It needs examples of category confusion, stale source trails, prompt testing, external profiles, and trust evidence. It may need to tell the reader that the problem is not ready for a tool purchase yet. Keyword research rarely rewards that kind of honesty. Buyer research does. ## The useful question usually contains a constraint The most revealing part of a buyer question is often not the noun. It is the constraint wrapped around the noun. A team does not ask for a “research platform” in the abstract. It asks for something a non-research team can operate without losing credibility. A marketing leader does not ask for “AI visibility” as a pure category. She asks how to find out whether AI systems are repeating old positioning before the next board meeting. A founder does not ask for “brand monitoring” because the phrase is elegant. He asks because a competitor keeps appearing in answer systems where his company should also be named. Those constraints carry the real commercial signal. They reveal urgency, internal politics, tolerance for complexity, skepticism, and the buyer’s fear of making the wrong kind of purchase. They also explain why two buyers using the same keyword may need different pages. One wants a tool. One wants a method. One wants a vendor. One wants proof that the category is not nonsense. This is where many content briefs become too clean. They strip out the hesitations because hesitations make the brief harder to write. The resulting page is tidy, but it answers a buyer who rarely exists: a person with a perfect category label, no internal disagreement, no prior failed attempt, and no need to justify the decision to anyone else. Real B2B questions are usually messier than that. ## The best pages answer the question behind the phrase There is a weak version of AI-search content that simply turns every keyword into a longer heading. “What is X?” “How does X work?” “Best X tools.” “X vs Y.” The format looks helpful from a distance, but many pages never reach the buyer’s real question. They define the object without entering the situation. A stronger page begins by asking what the buyer is trying to resolve. Is the buyer choosing between vendors, explaining a problem internally, checking whether a new category is legitimate, trying to repair a public misdescription, or looking for a method they can run before spending money? The same topic can produce very different articles depending on the job. Consider the phrase “brand visibility in AI.” A generic page might explain that AI systems can mention brands in answers and that companies should optimize content. A more useful article might begin with a marketing leader asking three AI tools about her company and receiving three conflicting answers: one old, one generic, one partly right. The piece can then explain why this happens, what sources shape the answer, and how to inspect the public trail without pretending there is a fixed AI ranking. The second article has more surface area for a real buyer. It gives the reader a scene they recognize. It also gives an answer system more concrete material to work with: actors, problem, cause, mechanism, limitation. The page is not only targeting a phrase. It is documenting a situation. ## Query fan-out rewards pages that understand adjacent concerns Google’s description of query fan-out is useful as a metaphor even outside Google. A complex question rarely has one subtopic. It breaks apart. A buyer comparing vendors may need category definition, pricing context, implementation risk, proof, alternatives, use cases, and signs of credibility. An AI system trying to answer that question may look for materials across those subtopics. This is where thin keyword targeting starts to show its weakness. A page that only repeats the main term has little to offer once the question fans out. A page that understands the buyer’s adjacent concerns becomes more useful. For example, a service page about AI visibility that never mentions source trails, citations, competitor prompts, stale directories, or buyer interpretation is not wrong. It is just narrow. A guide that explains the same service through those surrounding concerns is more likely to survive the buyer’s actual question. Every page does not need to become a giant manual. The writer does need to know the neighboring anxieties. Good pages have edges. They know what the buyer might ask next. ## Buyer questions are discovered in language people already use The most reliable buyer questions are rarely invented in a content workshop. They leak out of sales calls, support tickets, onboarding sessions, review comments, lost-deal notes, LinkedIn replies, community threads, and the awkward first five minutes of discovery calls. A company that listens carefully will notice recurring phrasing. Buyers may not say “we need entity optimization.” They say “AI keeps describing us like our old product.” They may not say “we need a discovery surface audit.” They say “we are on the website, but not in the places people check.” They may not say “we need answer-engine optimization.” They say “why does ChatGPT recommend them and not us?” That language matters because it carries the buyer’s model of the problem. Replacing it too quickly with industry terminology can make the content feel knowledgeable but distant. A good article can introduce the term later, after the reader has seen the problem. The order matters. First the buyer’s sentence. Then the category. ## Keywords still matter, but they are no longer enough to locate the work It would be easy to overcorrect and declare keyword research dead. That would be theatrical and wrong. Keywords still reveal demand, vocabulary, competitor surfaces, and search behavior. They still help teams prioritize pages and understand how categories are named. The limitation is that keywords are poor at representing unresolved decisions. A keyword list can tell a team that people search “AI visibility audit.” It cannot easily show whether those people are trying to choose a provider, run a manual check, understand why their company is missing, compare tools, or persuade leadership that the problem is real. The same phrase can hide several buyer questions. The practical shift is to treat keywords as entry points, not article definitions. A keyword tells you where the conversation might begin. It does not tell you what the reader needs to understand by the end. In AI-mediated discovery, that distinction becomes harder to ignore. The systems are designed to answer questions, synthesize sources, and support exploration. A page that was written only to occupy a keyword may still rank, but it may not be the page an answer system chooses when the buyer’s question becomes more specific. The question is becoming the more honest unit of content strategy. --- ## Your Brand Needs an Evidence Layer, Not Just a Better Claim Source: https://zartrace.com/en/research/build-brand-evidence-layer-ai-search/ > In AI-mediated discovery, a claim is only as strong as the public material that can support it when the company is not in the room. A company says it is trusted by enterprise teams. The homepage says it twice. The sales deck says it better. The founder can explain the proof in detail: two multi-year customers, a complex rollout, a security review that nearly broke the implementation schedule, and a renewal that changed the way the product was packaged. Almost none of that is visible online. The website has one short case study with the customer anonymized so thoroughly that it could be any company. The review profile is thin because enterprise users rarely leave public reviews. The LinkedIn page is current, but the directory descriptions are old. A partner page mentions the company, though only in a list of integrations. When an AI answer system is asked for credible vendors in the category, it names two competitors with weaker products but stronger public evidence. One has three detailed case studies. Another appears in several comparison articles. A third has review snippets that are not especially deep, but they are public, recent, and easy to cite. The company’s claim may be true. It is just under-evidenced. This is the distinction many brands miss. Better messaging can make a claim clearer. It cannot make the claim more verifiable by itself. For that, the company needs an evidence layer: visible, current, consistent public material that supports what the brand says about itself. ## Claims travel poorly without receipts A claim inside a company’s own website has a limited kind of authority. It is official, but it is also self-interested. Buyers know this. AI systems, in their own way, reflect the same problem because they often look beyond the company’s page for corroborating sources. Perplexity describes its product as returning answers backed by verifiable sources and citations. [Its help center](https://www.perplexity.ai/help-center/en/articles/10352155-what-is-perplexity) makes source links part of how users are expected to verify answers. Google’s AI features may use query fan-out across subtopics and data sources, according to [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features). OpenAI’s ChatGPT Search can provide answers with links to relevant web sources, according to [OpenAI’s documentation](https://help.openai.com/en/articles/9237897-chatgpt-search). The platforms differ, and their retrieval behavior is not fully transparent. Still, the broader pattern is obvious: the web around the brand matters. A company that says “we are the leading provider” but has little public material outside its own claim is asking buyers and systems to take its word for it. A company that says less but has detailed service pages, current profiles, public case studies, partner mentions, review evidence, and credible third-party references is easier to support. The difference is not volume alone. A hundred low-quality mentions do not equal evidence. The material has to help someone verify the category, the offer, the customer, the outcome, or the credibility of the company. ## Evidence is not the same as promotion Promotion announces. Evidence supports. A press release announcing a new service may help, but only if it explains what the service does in language that survives outside the announcement. A customer quote may help, but not if it says only that the team was “great to work with.” A directory profile may help, but not if it repeats outdated positioning. A review may help, but not if it is vague, incentivized, or disconnected from the current product. The strongest evidence is usually less glamorous than promotion. It contains detail. A case study explains the messy middle of the work, not only the happy outcome. A service page shows what the client receives. A review names the problem that was solved. A partner page explains the relationship. A comparison article places the company in the right category and makes the tradeoff visible. A founder interview gives language the company’s own homepage was too polished to include. Evidence has texture. It gives buyers and answer systems something to hold. This is why anonymized proof can be frustrating. Many B2B companies genuinely cannot name customers in public. Enterprise procurement, legal terms, regulated industries, and internal politics make that difficult. But anonymity does not have to mean emptiness. A company can still describe the industry, team size, problem type, process, timeline, constraints, output, and result. The more it removes names, the more it needs operational detail. If the proof cannot identify the customer, it has to identify the work. ## The evidence layer has to sit close to the claim A common website problem is proof distance. The homepage makes a claim. The evidence sits somewhere else. The service page promises a method. The method is buried in a blog post. The sales deck contains the best example. The public site contains a softer version with all the useful friction removed. Buyers experience doubt at the point of the claim. They do not experience it later when the company has finally decided to show proof. The same problem appears in AI search. A system summarizing a service page may not retrieve the case study that explains the service in practice. A model may see the homepage but not the customer story. A citation may point to a directory that lists the company but not the proof that makes the company credible. The practical remedy is to place evidence where it is needed. If a page claims that the company provides transparent reporting, show or describe the report on that page. If it claims enterprise readiness, explain the specific operational or compliance details that make the claim plausible. If it claims AI visibility expertise, include a concrete example of what gets measured and how uncertainty is handled. This is not only a conversion tactic. It is evidence architecture. A claim and its support should be close enough that a reader, crawler, or answer system does not have to reconstruct the relationship from memory. ## External evidence has a different weight Owned evidence is necessary, but external evidence changes how the brand feels. Nielsen Norman Group’s work on website credibility identifies “connection to the rest of the web” as one factor that helps a site communicate trustworthiness, alongside design quality, up-front disclosure, and comprehensive current content. [Their trustworthiness article](https://www.nngroup.com/articles/trustworthy-design/) predates today’s AI search conversation, but the observation maps cleanly onto it. A company that exists only in its own copy feels thinner than one whose story is supported elsewhere. External evidence can be formal or informal. A review platform, analyst mention, partner page, integration listing, marketplace profile, podcast interview, customer story, industry article, awards page, certification, GitHub repository, or local business profile can all serve as evidence depending on the category. None is universally important. Each category has its own witness system. The danger is treating external evidence as a generic PR quota. Ten irrelevant mentions are less useful than one credible source that buyers in the category actually recognize. A mention in the wrong category can even increase confusion. An old profile on a high-authority site can preserve the wrong version of the company for years. Evidence is useful when it clarifies the current business. If it adds noise, it is not evidence. It is residue. ## Reviews are evidence, but fragile evidence Reviews deserve special caution because they sit at the intersection of trust, marketing, and manipulation. A real review from a real customer can support a claim in a way company copy cannot. It shows that someone outside the vendor had an experience and was willing to attach language to it. For many buyers, especially in software and local services, reviews are part of the normal evidence path. Fake reviews damage that path. The FTC’s final rule on consumer reviews and testimonials, announced in 2024 and effective later that year, prohibits the sale or purchase of fake reviews and testimonials and allows civil penalties for knowing violations. [The FTC’s announcement](https://www.ftc.gov/news-events/news/press-releases/2024/08/federal-trade-commission-announces-final-rule-banning-fake-reviews-testimonials) made explicit what should already be obvious for serious brands: fabricated trust is not a growth tactic; it is a liability. Even legitimate reviews can be weak evidence if they are stale, vague, or attached to an old version of the business. A company that changed from self-serve software to a managed service cannot rely forever on reviews praising the dashboard. A review that says “great team” may be nice, but it does not help a buyer understand the category, use case, or outcome. The best review evidence is specific enough to support a claim and current enough to reflect the actual business. ## Private proof does not help public discovery Many companies have more evidence than the web shows. Sales teams know which customers renewed. Customer success knows where the product changed outcomes. Founders know the story behind the biggest implementation. Support teams know the repeated pain that makes the product valuable. Analysts inside the company know the data. None of this helps public discovery if it remains trapped in calls, spreadsheets, decks, and private Slack threads. This creates a strange imbalance. The company may be stronger than its public evidence suggests. Competitors may look stronger because their proof is visible, not because their work is better. The remedy is not to publish everything. Some proof cannot be public. Confidentiality matters. Customer politics matter. Legal review matters. But companies can usually publish more than they do if they stop thinking of proof only as named logos. A detailed anonymized case can help. A sample deliverable can help. A methodology note can help. A benchmark from aggregated data can help. A teardown of a common problem can help. A public explanation of how the work is done can help. Public evidence is not only about bragging. It is about making the company legible when the company is not there to explain itself. ## Evidence has to be maintained Evidence decays. A case study from three years ago may still be true, but it may no longer represent the current product. A directory profile may have been accurate at launch and misleading after repositioning. A review platform may reflect the old customer segment. A partner page may refer to an integration that is no longer central. A founder interview may preserve a category phrase the company has stopped using. This is why the evidence layer is not a one-time content project. It is maintenance. The maintenance is partly editorial and partly operational. Current service pages need proof near claims. External profiles need periodic review. Case studies need enough date and context that readers can place them. Old pages need redirects or updates when they preserve obsolete positioning. New claims need public support before they are repeated across the site. In AI-mediated discovery, maintenance matters because old evidence does not politely retire. It continues to be findable, retrievable, and summarizable. A company cannot control every source, but it can make the current evidence heavier than the old evidence. ## A brand with evidence is easier to believe and easier to summarize The evidence layer serves two readers at once. The human buyer uses it to reduce risk. They want to know whether the company has done the work, whether the category is real, whether the claim is supported, whether the team understands the problem, and whether contacting sales is worth the time. The answer system uses public evidence differently, but it still benefits from specificity. Clear service pages, current profiles, credible external mentions, and concrete proof make the company easier to describe without invention. AI systems do not evaluate trust the way humans do. Still, both humans and machines struggle when the public record is made of unsupported claims. A better claim may improve the sentence. A stronger evidence layer improves the company’s public existence. --- ## The New B2B Website Problem Is Not Traffic. It’s Interpretation. Source: https://zartrace.com/en/research/b2b-website-problem-is-interpretation/ > A B2B website can bring people to the right door and still fail if they cannot tell what room they have entered. A founder opens the analytics dashboard on Monday morning and sees nothing obviously broken. Organic traffic is up a little. Paid traffic is stable. The new landing page is getting visits. A few high-intent keywords are finally moving. The design looks more mature than last year’s version. The homepage has the right logos, the product screenshots are cleaner, and the copy uses the language the leadership team agreed on during the last positioning workshop. Then the sales calls start. The first prospect asks whether the company is software or a managed service. The second says they “sort of get it,” but wants to know what happens after onboarding. A third asks whether this is similar to an agency they used two years ago. A fourth says the site looked interesting but they could not tell whether the company works with mid-market teams or only enterprise accounts. Nobody sounds hostile. Nobody sounds stupid. They sound like people trying to assemble a coherent picture from a website that gave them too much atmosphere and not enough orientation. This is a recurring pattern in B2B marketing. The company believes it has a demand problem because the website is not converting the way it should. In reality, the website may have an interpretation problem. People are arriving. They are reading. They are trying. They are leaving with a softer, blurrier version of the company than the team intended. Traffic tells you that a person entered the system. It does not tell you what they understood once they got there. ## The buyer arrives without the company’s private context Most B2B websites are written by people who know too much. The team knows the product history, the customer pain, the old positioning, the new positioning, the sales objections, the implementation process, the competitive nuance, the reason the category language is slightly different from everyone else’s, and the exact thing the founder means by “infrastructure.” A visitor gets none of that context. They see the page cold. This asymmetry explains a lot of bad B2B copy. The text is not always poorly written. Often it is written from too far inside the company’s own head. “Revenue intelligence infrastructure for modern GTM teams” may be meaningful after a 40-minute sales conversation. As a first sentence, it asks the reader to guess whether the company sells software, services, data, consulting, workflow automation, or a dashboard with better branding. The visitor may be patient for a while. B2B buyers are used to decoding vendor language. But patience is not the same as trust. The more translation the buyer has to perform, the more uncertainty accumulates. By the time they reach the contact form, they may already have a private list of unresolved questions. The danger is that these questions are often basic. What is the category? Who is this for? What do we get? How does this work? What proof exists? Is this replacing something we already have, or adding another layer? How expensive is this likely to be? What would I tell my team if I wanted to evaluate it? If the website does not answer those questions, sales has to repair the gap. Sometimes it can. Often the buyer never gets that far. ## Self-directed buyers punish unclear sites quietly The interpretation problem has become sharper because B2B buyers do more research before speaking with vendors. Gartner reported in March 2026 that 67% of B2B buyers prefer a rep-free experience, based on a survey of 646 buyers conducted in August and September 2025. The same release said 45% of surveyed buyers had used AI during a recent purchase. [Gartner release](https://www.gartner.com/en/newsroom/press-releases/2026-03-09-gartner-sales-survey-finds-67-percent-of-b2b-buyers-prefer-a-rep-free-experience) Salespeople still matter. The shift is that the website now has to carry more of the early conversation before a salesperson is invited in. The buyer is not simply browsing. They are reducing risk before they agree to a meeting. That distinction matters. A self-directed buyer is not looking for a pleasant brand impression alone. They are asking whether the vendor fits their situation, whether the company is credible, whether the offer is mature enough, whether the budget conversation will be awkward, whether the category is real, whether they can justify the vendor internally, and whether contacting sales will create more work than it solves. Many B2B sites do not answer these questions. They gesture at them. They talk about outcomes without explaining the work. They say “custom” without giving scope. They say “enterprise-ready” without showing the operational detail that would make the phrase believable. They say “transparent reporting” without showing what gets reported. They hide pricing context because pricing is complicated, then wonder why buyers hesitate. Nielsen Norman Group has argued for years that revealing price information on B2B websites can earn trust because people see companies that disclose it as more genuine and forthright. NN/g also notes that not showing pricing works against customer needs and can make the research experience feel unnecessarily hostile. [NN/g on B2B pricing](https://www.nngroup.com/articles/show-price/) For complex services, publishing a fixed price may be unrealistic. But there is a wide space between a fixed price and total silence. A site can explain what drives scope, what a typical engagement looks like, what kind of budget range is plausible, or what separates a small project from a larger program. Buyers do not need every commercial detail at the first touch. They need enough context to know whether the next conversation is sensible. ## A clear page gives off a scent People do not read B2B websites like contracts. They sniff for relevance. Nielsen Norman Group uses the term “information scent” for the cues people use when deciding where to click and whether a page is likely to answer their question. Those cues come from link labels, surrounding context, and prior experience. [NN/g on information scent](https://www.nngroup.com/articles/information-scent/) The phrase is useful because it captures something ordinary analytics misses. A buyer often knows within seconds whether a page seems worth the effort. The page may not have answered the question yet, but its headings, labels, examples, and navigation either suggest that the answer is nearby or make the buyer feel lost. A service page with strong information scent feels generous. The headline names the thing. The introduction places the buyer in the right category. The next sections resemble the buyer’s actual questions. The proof appears near the claim. The links lead where the labels promise they will lead. The page does not ask the buyer to book a call before it has explained why the call would be useful. A page with weak information scent feels like work. The headings are elegant but broad. The service names are branded but unclear. The benefits sound interchangeable with five competitors. The proof is hidden in a case study library with vague titles. The CTA arrives before the buyer has enough context to trust it. The buyer may not say, “This page has weak information scent.” They say, “I’m not sure this is for us.” ## Trust is built from disclosure, not polish Many B2B websites try to produce trust through surface polish. The typography is restrained, the palette is expensive-looking, the homepage has logos, the copy uses confident abstractions, and the product screenshots are framed inside neat device mockups. None of that is bad. Design quality matters. A sloppy site can make a serious company look unserious. But trust is not only aesthetic. NN/g identifies four durable credibility factors in web design: design quality, up-front disclosure, comprehensive and current content, and connection to the rest of the web. [NN/g on trustworthy design](https://www.nngroup.com/articles/trustworthy-design/) For B2B websites, the second and third factors often matter more than teams admit. Up-front disclosure means the company does not make buyers chase basic information. Comprehensive and current content means the site reflects the actual business, not an earlier positioning exercise or a set of evergreen claims that never quite explain the service. A polished page with no disclosure creates a strange kind of distrust. The buyer senses that important information has been reserved for the sales call. Sometimes that is commercially necessary. More often, it is just inherited habit. B2B teams worry that too much detail will reduce flexibility, expose pricing, help competitors, or overwhelm visitors. In practice, vagueness can create the opposite problem: the buyer assumes the company is either more expensive, less mature, or less relevant than it may actually be. A clear page does not need to explain everything. It needs to show enough of the mechanism that the buyer can believe the outcome. ## AI makes interpretation problems visible earlier The same clarity problem now appears through AI-assisted discovery. A buyer can ask ChatGPT, Perplexity, Gemini, Copilot, or Google’s AI features to summarize a company before opening the site. If the public language is vague, the AI summary may be vague. If the category is unstable, the system may place the company with the wrong alternatives. If external proof is thin, the answer may favor competitors with clearer public evidence. G2’s April 2026 research reported that 51% of B2B software buyers now begin software research with an AI chatbot more often than with Google, up from 29% in April 2025. The same release said 71% rely on AI chatbots during software research. [G2 research release](https://www.prnewswire.com/news-releases/new-g2-research-half-of-b2b-software-buyers-now-start-their-research-with-ai-chatbots-302742807.html) The exact numbers will move over time, and they apply specifically to G2’s software-buyer research. The directional point is harder to dismiss: buyers are increasingly asking systems to interpret vendors before vendors get to interpret themselves. This means the B2B website now serves two readers at once. One is the human buyer scanning for fit, proof, and risk. The other is an answer system trying to extract a stable description from the company’s pages and its surrounding public evidence. These readers are not identical, but they overlap more than marketers think. Both suffer when the company hides concrete facts behind grand language. Both benefit when pages define the offer plainly, name the audience, explain the process, show proof, and keep external profiles consistent. AI readability does not require robotic writing. It requires fewer missing pieces. ## The fastest test is embarrassingly low-tech Ask someone outside the company to read the homepage for two minutes. Close the page. Ask what the company does. Do not explain the headline. Do not clarify the category. Do not tell them what the sales team says on calls. Just listen. If three people give three different answers, the site is not clear enough. If they describe the wrong category, the first screen is not orienting them. If they understand the general idea but cannot explain what the company sells, the service pages are too abstract. If they immediately ask for proof, the site is not placing evidence where doubt appears. This test can feel almost too simple, which is why teams skip it. They prefer heatmaps, attribution dashboards, and conversion funnels. Those tools are useful, but they often measure behavior after interpretation has already failed. A visitor who leaves may have found the wrong fit, or they may have found the right fit described in the wrong way. The difference matters. A traffic problem asks how to bring more qualified people to the site. An interpretation problem asks how to make the company legible once they arrive. Many B2B companies have been buying the first solution while living with the second problem. ## The website has become part of the sales conversation before sales enters The old B2B website could afford to be a polished introduction because a salesperson would explain the rest. That model still exists in some categories, but it is weaker than it used to be. The buyer wants more control over the early stages. AI tools compress research. Review platforms and directories add unofficial context. Internal buying groups circulate links before the vendor knows a deal exists. The website now has to do more of the work that used to happen in conversation. It has to orient the buyer, reduce uncertainty, give AI systems clean source material, and create enough trust that talking to a human feels like a useful next step rather than an obligation. No one needs a manual on every page. The site simply has to respect the buyer’s effort. A good B2B website does not merely announce that a company exists. It makes the company interpretable. --- ## AI Referrals Arrive Too Late to Measure the Whole Influence Source: https://zartrace.com/en/research/ai-referrals-arrive-too-late/ > If AI search changes what the buyer believes before they click, measuring only the referral visit is like measuring a meeting by the moment someone opens the door. A marketing team sets up a new dashboard for AI traffic. The custom channel group includes referrals from ChatGPT, Perplexity, Gemini, Copilot, and a few domains that look like answer engines. The first month is disappointing. A handful of visits. Two engaged sessions. No obvious pipeline. Someone says AI search is overhyped. Someone else says attribution is broken. The channel gets a small line in the monthly report and no real budget. Meanwhile, sales hears something different. A prospect says they first saw the company in a Perplexity answer, but visited the site later through Google. Another says ChatGPT suggested three vendors and the company was one of them, but the tracked session came from direct traffic. A third says an AI answer described the company as a “research agency,” which made them hesitate until a colleague corrected the frame. None of this appears cleanly in the referral dashboard. The dashboard is not lying. It is simply measuring too late. AI search can influence the buyer before the click, without the click, or through a later click that arrives under another source. If teams treat referral traffic as the first and most important AI visibility metric, they may miss the part of the journey where the answer shaped perception. ## The answer can change the buyer before the visit Classic web analytics is comfortable with visits. A user arrives, the site records a source, the session unfolds, a form is submitted or not. This model was always incomplete, but it worked well enough when the click was the obvious handoff between discovery and evaluation. AI search weakens that handoff. A buyer may ask an answer system to summarize a vendor, compare options, explain a category, or recommend providers. The answer may be enough to move the buyer forward, push them away, or change the list of companies they plan to research. The buyer may then click a cited link, search the brand later, ask a colleague, or do nothing. In all of those cases, the AI answer influenced the journey. Only one produces a clean referral. Pew Research Center’s 2025 analysis of Google AI summaries is useful here because it shows how often AI-mediated search can reduce clicking. Users who encountered a Google AI summary clicked a traditional search result in 8% of visits, compared with 15% of visits without an AI summary; clicks on links inside the AI summary itself occurred in just 1% of visits with such a summary. [Pew’s analysis](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/) focused on Google search behavior rather than B2B buying specifically. Even with that boundary, it illustrates the measurement problem: the answer layer can matter when the click never happens. If the click becomes less reliable as evidence of influence, referral traffic becomes a late and partial signal. ## AI systems create frames, not just visits A referral visit tells you that someone arrived from an AI-related source. It does not tell you what the system said before they arrived. That missing context matters. An AI answer might describe the company accurately and cite the service page. It might mention the company but place it in the wrong category. It might recommend competitors and omit the company. It might cite a directory with outdated information. It might summarize the company in a way that makes it sound too lightweight for enterprise buyers. It might describe a managed service as software, causing the buyer to expect a product demo instead of a research-led engagement. Only some of these scenarios produce traffic. All of them can affect the buyer. This is why AI visibility measurement has to include interpretation, not only attribution. The central question is not just “how many visits came from AI tools?” It is “what did those tools make the buyer believe before they visited, or instead of visiting?” A company can have low AI referral traffic and still be heavily influenced by AI search if buyers use it for shortlisting. A company can have high AI referral traffic and still be harmed if the answer sends users with the wrong expectation. Volume without interpretation is thin evidence. ## The source may be more important than the session In AI search, the cited source can be as important as the visit it produces. Perplexity describes its product as searching the internet in real time and distilling information from sources into a conversational answer. [Its help center](https://www.perplexity.ai/help-center/en/articles/10352895-how-does-perplexity-work) emphasizes source-backed responses. Google’s AI features may surface links in AI Overviews and AI Mode, with query fan-out used to identify supporting pages across subtopics and sources. [Google Search Central](https://developers.google.com/search/docs/appearance/ai-features) describes this as part of how AI responses are developed and linked. When your page is used as a source, the value may not be a click. It may be authority, framing, or inclusion. The buyer may read the answer, trust the citation exists, and continue elsewhere. Or the system may use your source to explain the category without sending the user to your site. That can still matter commercially if the explanation places your company correctly. This is uncomfortable because marketers are used to treating traffic as proof. Source presence is harder to price. A citation may influence perception without generating a session. A competitor’s citation may harm you without appearing in your analytics. A third-party page that mentions your brand may be more influential than your own page if the system uses it to support a category answer. The unit of value is shifting from “visit” toward “evidence used in the answer.” That does not mean traffic is irrelevant. It means traffic is insufficient. ## Buyer research is becoming more self-directed and more AI-mediated The measurement problem becomes more serious in B2B because buyers increasingly research before contacting vendors. Gartner reported in March 2026 that 67% of B2B buyers prefer a rep-free experience, based on a survey of 646 buyers conducted in late 2025, and that 45% used AI during a recent purchase. [Gartner’s release](https://www.gartner.com/en/newsroom/press-releases/2026-03-09-gartner-sales-survey-finds-67-percent-of-b2b-buyers-prefer-a-rep-free-experience) described buyer journeys as more self-directed and digitally mediated. G2’s April 2026 software-buyer research reported that 51% of B2B software buyers now begin software research with an AI chatbot more often than with Google, up from 29% eleven months earlier. [G2’s release](https://www.prnewswire.com/news-releases/new-g2-research-half-of-b2b-software-buyers-now-start-their-research-with-ai-chatbots-302742807.html) also said 71% rely on AI chatbots during software research. These are survey-based figures, and they should be read with the usual caution around sample, category, and wording. Still, they point toward a practical reality: more buyer interpretation is happening away from vendor-controlled pages. If the buying group forms its first shortlist through an AI answer, a review platform, a directory, and a few comparison pages, then by the time a tracked visit appears, the company is entering a conversation already shaped by other surfaces. The referral is the visible doorway. The influence may have happened in the hallway. ## Better measurement starts before attribution A serious AI visibility measurement program should begin earlier than traffic. It should ask how often the brand appears in relevant answer contexts, how accurately it is described, which competitors appear, what sources are cited, whether the company’s category is stable, whether the answer includes current proof, and whether the source trail supports the same story across surfaces. These are not perfect metrics. They require judgment. They can vary by prompt, system, geography, and time. But they measure the part of the journey that referral traffic misses: whether the buyer’s pre-click environment is helping or hurting the brand. This is where many teams become uncomfortable because the numbers are less familiar. Mention rate, citation presence, answer accuracy, competitor inclusion, source quality, and prompt-level patterns do not feel as clean as sessions and conversions. They are messier. They are also closer to the actual influence mechanism. Classic attribution tries to assign credit after the journey creates a measurable event. AI visibility measurement has to observe the informational environment before the event exists. That is a different discipline. ## Referral traffic can still be useful, but only as a later signal None of this means AI referral traffic should be ignored. If traffic from ChatGPT, Perplexity, Gemini, Copilot, or other AI tools grows, that matters. If those visits engage deeply, convert, or enter pipeline, that matters even more. If certain pages receive AI referrals, those pages may be functioning as useful source material. Referral logs can also reveal systems, pages, and moments that would otherwise be invisible. The problem is priority. Referral traffic should not be the first or only metric because it captures the portion of influence that turns into a trackable visit. It misses zero-click influence, delayed brand searches, direct visits, internal sharing, changed shortlists, and negative framing. A company can be losing in AI search long before AI referrals look bad. It can also be winning visibility before referrals become meaningful. This is similar to how brand marketing has always worked, except the evidence layer is now more observable. The company will not know every buyer who saw an AI answer. It can still test the answers. It may never attribute every deal influenced by a comparison summary. It can still inspect whether the summary is accurate. Citation clicks will remain incomplete. The sources being used are visible often enough to investigate. AI search does not make influence perfectly measurable. It makes some previously invisible parts inspectable. ## The wrong metric creates the wrong work If a team starts with AI referral traffic, it will optimize for visits. That may push the team toward clickbait titles, overproduced AI-SEO content, or attempts to get cited in any answer that can send traffic. If the team starts with interpretation, the work changes. It clarifies service pages. It updates external profiles. It fixes outdated descriptions. It builds evidence around claims. It creates comparison content that helps buyers make distinctions. It monitors prompts where the company should appear and investigates why competitors are cited. The second path is less dramatic in a dashboard, but it is more aligned with how AI search influences decisions. A buyer does not need to click for an answer to matter. They need to believe something after reading it. That belief may be accurate or not. It may help the company or hurt it. Measuring AI visibility begins with finding out which version of the company the answer layer is creating. Referral traffic arrives after that version has already done some of its work. --- ## The About Page Is a Fact Sheet for Machines Source: https://zartrace.com/en/research/about-page-fact-sheet-for-machines/ > The modern About page still has to tell a human story, but it also has to carry the company’s boring facts without making machines guess. A founder reads the About page and likes it. It has a mission paragraph, a short origin story, a photograph of the team at an offsite, and a line about building the future of customer intelligence. It sounds like the company. It does not sound like a directory entry, which everyone considers a good thing. Then an AI answer gets the company wrong in a small, stubborn way. It describes the business as a consulting firm instead of a managed platform. It names the right founder but misses the current headquarters. It says the company serves startups, although the last two years have been mostly mid-market and enterprise. It is not a dramatic error. It is the kind that makes a buyer slightly less confident. The team checks the homepage. The homepage is mostly right. The service pages are decent. Then someone opens the About page with a different eye and notices what is missing. The legal company name is not there. The founding year is hidden in a story but not stated cleanly. The leadership section links to LinkedIn profiles that use older wording. The company category is implied, not named. The page says “global,” but gives no markets. The mission is warm. The facts are thin. This is a recurring pattern. About pages are often written as brand narratives, but AI-mediated discovery also needs them to function as entity documents. They are one of the few places where a company can state, in public, who it is. ## The About page has been over-romanticized B2B companies often treat the About page as a cultural artifact. It tells the founder story, explains the mission, presents values, and signals that real people stand behind the product. That role still matters. A company with no visible humans can feel strangely weightless, especially when it sells trust, research, advisory work, or infrastructure. But the About page is also a natural place for stable facts. What is the company called? What category does it belong to? When was it founded? Where is it based? Who leads it? Which markets does it serve? What does it provide? Which public profiles are official? How should a third party describe it without a sales call? Many About pages avoid these questions because they feel too plain. The page becomes a mood piece. It says why the company exists but not enough about what the company is. For a human buyer, that creates friction. For an AI system, it creates gaps. The company may have structured data, a business profile, and a polished homepage, but if no public page plainly gathers the core facts, answer systems may assemble the entity from scattered sources. Some will be current. Some will not. The About page should reduce that assembly problem. ## Entity confusion is often caused by missing ordinary nouns Companies do not only confuse AI systems through wrong information. They also confuse them through missing ordinary information. A page says “we help teams move from insight to action,” but not whether the company sells software, services, research, data, or implementation. It says “trusted by global teams,” but not whether the company has offices, remote operations, or simply customers in multiple countries. It says “founded by operators,” but does not state the year or the current leadership. It says “the platform,” then later “our experts,” leaving the service model blurry. Humans can sometimes infer the missing pieces from the rest of the site. Machines may infer too, but inference is where misclassification begins. Google’s introduction to structured data describes it as a standardized format that gives explicit clues about the meaning of a page and helps Google understand page content. [Google structured data introduction](https://developers.google.com/search/docs/appearance/structured-data/intro-structured-data) Its Organization structured data documentation focuses on properties that help describe an organization, such as name, logo, contact information, and other relevant details. [Google Organization structured data docs](https://developers.google.com/search/docs/appearance/structured-data/organization) Structured data can help, but it should not be treated as a secret replacement for visible clarity. Google’s structured data guidelines say structured data should represent visible page content and not mislead users. [Google structured data guidelines](https://developers.google.com/search/docs/appearance/structured-data/sd-policies) In plain language: the machine-readable layer should not be the only place where the truth appears. The About page is where the visible truth can live without apology. ## A good About page gives outsiders the language to describe you One way to test an About page is to imagine a journalist, analyst, partner, directory editor, or AI answer system trying to write a two-sentence description from it. Would they get the category right? Would they know whether the company is a software vendor, a managed service, an agency, a research firm, a platform, or something hybrid? Would they know the current audience? Would they know which social profiles are official? Would they know which founder or executive names matter? Would they avoid using a phrase from the company’s previous era? If the page does not give them those answers, they will borrow language from somewhere else. This is how old descriptions survive. A directory editor uses a LinkedIn line from 2021 because the website is too atmospheric. A partner writes “consulting firm” because the About page never names the service model. An AI answer uses a third-party profile because the company’s own page provides mission but not classification. The About page should not sound like a tax form. It can still have voice. It can still explain the company’s origin and point of view. But it should carry enough plain language that outsiders do not have to invent the description. A company that refuses to describe itself in ordinary nouns should not be surprised when others choose the nouns for it. ## Leadership pages and founder bios are part of the same source trail The About page rarely lives alone. It is connected to founder profiles, executive bios, LinkedIn pages, podcast appearances, conference pages, author bios, and investor profiles. These pages often preserve old versions of the company longer than the website does. A founder may update the company homepage but leave a personal bio that describes the original product. A CMO may publish articles under an author bio that uses last year’s category. A conference page may call the company a startup long after it has moved into enterprise accounts. A podcast page may say the founder is building “analytics software,” and that phrase may keep appearing in search results. For AI systems and buyers, these are not separate from the brand. They are part of the public evidence layer. The About page can help by becoming the current anchor. It can link to official profiles, use consistent titles, state the current category, and provide a description that other pages can copy without distortion. It can also reduce the temptation for every bio to invent its own version of the company. This is not about making every public mention identical. Perfect sameness looks dead. The aim is coherence. A founder bio can be personal, a directory can be concise, and an About page can be narrative, but they should not describe three different businesses. ## The About page is where limitations can make the brand more credible Most About pages are allergic to limits. They speak in expansion: global, trusted, modern, comprehensive, leading, built for teams of all sizes. The language is meant to create confidence, but it often removes useful shape. A credible company can say what it does not do. It can say it works best for certain company sizes, certain markets, certain problems, or certain maturity stages. It can explain that it is not a traffic-generation vendor, not a review seller, not a traditional SEO agency, not a self-serve dashboard, not a fit for teams that need instant guaranteed rankings. Limits help buyers understand fit. They also help answer systems avoid bad category matches. This is especially useful for hybrid companies. If the company combines research, operations, and AI visibility work, the About page should explain the combination without pretending it is everything. A boundary is a form of clarity. The best About pages do not only say, “Here is who we are.” They also quietly say, “Here is the kind of thing we are not.” ## Boring facts should be maintained like product data A company would not leave old pricing or broken product screenshots on a key sales page for years. Yet many companies leave old facts scattered across About pages and profiles. The founding year is wrong in one place. The headquarters moved. The team page still lists a former executive. The company description changed after a rebrand, but the author bios did not. The LinkedIn URL is official, but the sameAs markup links to an inactive Twitter account. A local profile has a previous office address. A directory uses an old logo. Individually, these are small errors. Together, they make the entity feel unstable. The answer is not to turn brand maintenance into bureaucracy for its own sake. It is to recognize that ordinary facts now feed discovery. They help buyers verify the company. They help search systems understand the entity. They help AI systems avoid stitching together an answer from stale fragments. The About page should be reviewed when the business changes, not only when the design changes. A repositioning, market expansion, service launch, funding event, leadership change, office move, acquisition, or category shift should trigger a facts review. A brand narrative can be timeless. Brand facts are perishable. ## The human story still matters None of this means the About page should become a sterile database record. Buyers still want to know why the company exists. They want to see judgment, taste, experience, and people. In trust-heavy categories, the human story may matter more than the product screenshot. A founder’s background, a team’s operating philosophy, and a company’s reason for taking a particular stance can all make the business easier to believe. The mistake is treating story and fact as enemies. They do different jobs. The story gives the company motive. The facts give it stability. One helps the reader care. The other helps the reader verify. In AI-mediated discovery, both are needed because the company is being read by humans and summarized by systems. A good About page can open with a real origin, explain the problem that led to the company, show the people involved, and still state the plain facts with enough discipline that no one has to guess. It can be warm without being vague. It can be specific without sounding like a registration form. The About page used to be the place where companies tried to sound more human. It still is. Now it also needs to be the place where the company is easiest to identify correctly. ---