Five out of thirty-six. That is how many times a mid-market ERP vendor's name showed up when I ran twelve comparison prompts through three AI assistants on their behalf — a composite scenario built from patterns I see across the audits I run, not a single named client, which I will label clearly as we go. Fourteen percent. And this was a company that ranks on page one of Google for nearly every relevant term in its category.
The tension is not complicated once you see it. Ranking well on Google measures whether your own website can answer a query. It says nothing about whether you appear in an answer that an AI assistant has built almost entirely out of somebody else's page.
The Query Type Your Website Was Never Built to Answer
Most of the work brands have done to be visible in AI search has focused, understandably, on general informational queries — "what does an ERP system do," "how does cloud accounting work," "what should a mid-market manufacturer look for in inventory software." For queries like these, an assistant often synthesizes across a range of sources, and a well-structured, well-cited page on your own site has a real shot at being one of them.
Comparison and shortlist queries behave differently. When a buyer asks "best ERP for a 200-person manufacturer" or "Brand A vs Brand B," the assistant is very rarely reasoning from first principles across the category. It is looking for a document that has already made this comparison, and it paraphrases that document. If a third-party page, a review directory, or a forum thread has already ranked or compared the options, the assistant tends to lean heavily on that existing structure rather than building its own from scratch.
This is a distinct behavior, and it means a brand can be strong on informational queries and functionally invisible on comparison queries at the same time. Nobody has optimized for it, because until recently nobody had a reason to look at comparison answers as their own category of AI-search behavior. Most content calendars still treat "best X" content as an SEO listicle problem rather than as the exact document an assistant will later summarize word for word.
A Composite Worked Example: 36 Queries, One Shortlist Gap
Here is the scenario, laid out step by step, so the mechanism is concrete rather than abstract. To be clear: this is a composite, illustrative example built from patterns across multiple engagements, not a report on one identifiable company.
The exercise started with twelve realistic comparison prompts a genuine buyer might type — things like "best ERP for a mid-size manufacturer," "the category's leading vendor vs this one," and "top ERP systems for multi-currency operations." Each of those twelve prompts was run against three widely used AI assistants, for a total of thirty-six queries.
The vendor's name appeared in five of those thirty-six answers. Fourteen percent. In the other thirty-one, a competitor was named instead, or the vendor was omitted from the shortlist entirely.
The next step was tracing where those thirty-one answers actually came from — not guessing, but reading each response for the structure and phrasing of an existing document. Of the thirty-one misses, twenty-four traced back to the same four places: three third-party pages — a regional software-review directory, a global software-comparison directory, and a category roundup published by an industry publication — plus one recurring, well-upvoted community thread that ranked the same handful of vendors in almost the same order every time it was cited. The remaining seven answers did not point cleanly to a single traceable source.
Two things followed from that finding. First, the vendor reached out to the three pages and corrected or added their listing — updated positioning, accurate feature detail, a fair representation next to the names the assistants already trusted. Second, they published one comparison page of their own, structured around four buyer decision criteria their actual prospects cared about, rather than a generic feature table.
Three weeks later, the same twelve prompts were run again across the same three assistants. The vendor was named in nineteen of thirty-six answers — fifty-three percent, up from fourteen. Nothing about the vendor's underlying product had changed in three weeks. What changed was whether the assistants had a source to quote from that included them.
The Four Places Comparison Answers Actually Come From
Across the audits I run, comparison answers tend to trace back to a short, repeatable list of source types. Knowing this list turns a vague worry — "we don't show up" — into a findable, fixable list of places to check.
- Third-party review and comparison directories — the G2-style platforms and category roundup sites that already rank or compare vendors in your space, whether or not you have ever claimed a profile there.
- Competitor-authored "us vs. them" pages — pages a rival has written specifically to compare themselves to you. Assistants cite these even when the framing is unflattering to you, because the structure is exactly what a comparison query is looking for.
- Community threads — Reddit discussions, forum posts, and niche community recaps that assistants increasingly surface, particularly for buyer categories where practitioners trade candid opinions.
- Analyst and press roundups — "top ten" or "best of" articles from trade press and industry analysts, which read as authoritative and get cited disproportionately relative to how often they are actually updated.
Almost none of these are pages a brand controls by default. That is precisely why being named on them requires deliberate effort rather than good SEO hygiene on your own domain.
Closing the Gap: Get Named, Then Get Ahead
The fix has two parts, and the order matters.
The first is corrective: identify the specific pages an assistant is actually quoting from — not a guess at where you think you should appear, but the pages that show up when you trace real answers — and pursue accurate inclusion on them. Sometimes that is a claimed profile update. Sometimes it is a factual correction to a comparison that has gone stale. Sometimes the page owner has no incentive to add you, and the request goes nowhere.
The second part does not depend on anyone else saying yes. Publish your own comparison content, built around the criteria your buyers actually use to decide — not a feature checklist, but the two or three trade-offs a real buyer weighs when choosing between you and the alternatives. A page structured this way is written in the shape an assistant is already looking for when it fields a comparison query, which makes it citable on its own terms, independent of whether the third-party directories ever add you.
Both moves matter, but only the second one is entirely within your control — which is why it is the one most worth doing regardless of how the outreach goes.
If you want to see where your own category stands, the test above is simple enough to run yourself: pick a realistic set of comparison prompts, run them across two or three assistants, and read each answer for whose structure it is borrowing. If you would rather have a second pair of eyes on the results, or want to talk through what a GEO audit for your category would look like, you can reach me through jonahtebaa.com.
For more on this and related work, see BrianServes, the platform for deploying autonomous AI e-mployees and Webspot, the AI strategy firm in Beirut.
Related evidence: Google itself treats search ranking and AI use of content as separate systems: its crawler documentation says the Google-Extended token governs whether crawled content may be used to train and ground its Gemini models, and that it does not affect a site's inclusion or ranking in Google Search. (Google's documentation on the Google-Extended crawler token)
A 2023 arXiv paper introduces Generative Engine Optimization as a framework intended to help content creators improve how visible their content is in the responses generated by generative search engines. (the arXiv paper introducing Generative Engine Optimization)
