# When AI Visibility Goes Up and Deals Go Down: Tracking What AI Says About Your Brand

URL: https://whatmodelsays.com/track-ai-perception-not-just-mention/
Published: 2026-10-16

> Visibility shows you were in the answer. Buyers decide from what the answer says about you. Pick the perceptions you want AI to hold and the objections you want gone, then check over time whether answers agree with each one.

You present the quarterly numbers: AI share of voice up 38%, mentioned in 9 of 10 category answers. The board nods. Two weeks later, sales flags three lost deals. In each one, the prospect said some version of "ChatGPT said you don't integrate with HubSpot." You shipped that integration months ago. Every one of those answers counted toward your 38%.

Being mentioned isn't being chosen. Write down the few perceptions you want answer engines to hold, and the objections you want gone, like "doesn't integrate with HubSpot." Then check, over time, whether the answers hold each perception and drop each objection.

By answer engines we mean AI assistants that answer questions directly, such as [ChatGPT](https://chatgpt.com), [Claude](https://claude.ai), [Gemini](https://gemini.google.com) and [Google AI Overviews](https://search.google/ways-to-search/ai-overviews/). A perception is any claim they make about your brand, such as "expensive" or "best for small teams."

We're the team at [Genezio](https://genezio.com), and we built a feature called Monitor perceptions for this job. But the method doesn't depend on our tool: ask the same questions on a schedule, and judge each answer against each claim.

## Why isn't a visibility score enough?

Visibility counts appearances, but the choice happens after you appear. In our [brand visibility definition](https://genezio.com/docs/core-concepts/brand-visibility), visibility is the share of eligible conversations where your brand appears. "If the brand shows at any location in the answer," it counts.

Recommendation rate works differently. Our [brand recommendation definition](https://genezio.com/docs/core-concepts/brand-recommendation) divides conversations where you're recommended by conversations where you're visible. It "measures your conversion rate from visibility to recommendation."

**What this means:** the gap between mentioned and recommended sits entirely inside answers that already name you. A mention count can't see into that gap. To explain it, you have to read what those answers say about you.

Mention studies can leave that part out. [SparkToro](https://sparktoro.com)'s large [consistency study](https://sparktoro.com/blog/new-research-ais-are-highly-inconsistent-when-recommending-brands-or-products-marketers-should-take-care-when-tracking-ai-visibility/) recorded brand lists and said so openly: "we didn't even try to collect data on how the AIs described each brand." (We've covered [mentions versus recommendations](https://whatmodelsays.com/what-is-geo-technical-definition-2/) before.) The next question is whether the description reaches buyers at all.

## Does what AI says about you reach buyers?

Buyers say it does. [G2](https://www.g2.com)'s [Answer Economy survey](https://company.g2.com/news/g2-research-the-answer-economy) asked 1,076 B2B (business-to-business) software buyers in March 2026. It found AI chatbots are now "the top source influencing buyer shortlists, ahead of review sites, analyst firms, and vendor websites."

The same survey reports that "69% of buyers chose a different vendor than initially planned, simply because it was part of the chatbot's recommendation." That's self-reported, and G2 sells into this market.

**The implication:** AI answers now shape who makes the shortlist. If an answer describes you as the wrong fit, you can be out before a buyer ever books a demo or opens a free trial. So the question isn't only whether AI mentions you. It's whether AI describes you the way you want buyers to see you.

## Does how AI describes you change whether it recommends you?

Yes: models recommend from the product information in front of them. In one [controlled test](https://arxiv.org/abs/2606.17443), researchers gave three models the specs, ratings and marketing copy for competing skincare products. When the specs were identical, the well-known brand won. Give a competitor a slightly better rating, and that edge disappeared.

The authors' conclusion is blunt: "The barrier for new brands is not brand equity itself, but the lack of any differentiating information."

We know that live answers work the same way. A buyer who asks for "easy to set up" and reads "hard to set up" next to your name is being steered toward someone else. One damaging claim, placed exactly where the buyer is unsure, can outweigh many neutral mentions.

**What this means:** what answers say about you deserves the same tracking as whether they mention you. That's the job Monitor perceptions does, and it starts with choosing which perceptions to watch.

## Which perceptions should you monitor?

Two kinds. The first is an objection you want gone, such as "expensive" or "hard to set up." Find these in your lost-deal notes.

The second is a position you want to earn, such as "best for small teams" or "suitable for regulated industries." Take these from your positioning doc. For this kind, our docs say "you want to know when answer engines start to say this without help."

Unsure where to start? Genezio also extracts what answer engines already say about your brand. Until you monitor a perception, the Monitor perceptions page shows "the six claims that answer engines state most often about your brand." Use them as inspiration: track one as is, or rewrite it into the claim you actually want buyers to hear.

## How does a perception check work?

A check gives you a verdict: does a claim agree with what answer engines say about your brand right now? For a perception you want, like "easy to set up," you want agreement. For an objection you want gone, like "doesn't integrate with HubSpot," you want disagreement. Every check is kept, so over time you build a history of what answer engines said and when it changed.

The answers come from conversations Genezio runs with answer engines. Each [topic](https://genezio.com/docs/core-concepts/topics) is a subject area buyers ask about, such as "CRM for startups." Most topics have a [persona](https://genezio.com/docs/core-concepts/personas): a typical buyer with a language and a location. Genezio acts as that persona and writes the prompts and follow-up questions the way that buyer would.

A monitor uses your report's filters, such as topics and models. Each check reads only the answers that came in since the last check, and turns them into one verdict.

Why not just read a few answers yourself? Because one answer tells you very little. Answer engines are non-deterministic: ask the same question twice and you can get two different answers, which is exactly what SparkToro's study found at scale. And you can't realistically read hundreds of answers by hand every week. That's why Genezio runs many conversations and condenses the new answers into a single verdict.

To start, open Monitor perceptions for your brand. Pick a suggested perception or write your own. Then come back after your next content release, or select Refresh to run a check now.

## The playbook

With the check in place, here's how to put it to work:

1. Write down 3 to 5 perceptions you want answer engines to hold, taken from your positioning.
2. Add the 2 or 3 objections that come up most in lost deals, plus any outdated facts, like a feature you've shipped that answers still say you lack.
3. Put each one under a monitor, scoped to the topics where buyers choose, such as comparisons and "best X for Y" questions.
4. Log the date of every release, and re-check after each one.
5. Read the verdict history next to your visibility and recommendation rate, not instead of them.
6. When a damaging claim persists or returns, open the cited sources.
7. Fix it at the source: update your own pages, and correct or answer the third-party pages being cited.

Visibility tells you that you're in the conversation. Perceptions tell you what buyers hear when you are. Pick the one perception you most want buyers to hear, and [put it under observation](https://genezio.com/docs/insights/monitor-perceptions).