Do Google Reviews Decide If ChatGPT Recommends You?

Sometimes, and the “sometimes” is the whole answer. When ChatGPT answers a local buyer question with a business panel, that panel is assembled from Google Business Profile data and ordered largely by rating and review volume — content cannot qualify you for it. When it answers the same question in prose, it builds the shortlist from web pages, and reviews become one signal among several. Which format you get depends on how the question is phrased, and it changes week to week.

The stakes are easy to underestimate. Research from Trustpilot, reported by TechRadar Pro (2026), found that businesses with 80 or more reviews appeared in over 75% of AI answers, against roughly 1% for businesses with none. That is not a ranking penalty. It is a qualification threshold.

What does the evidence actually show?

Across nine weeks of sampling the same buyer question, the format of ChatGPT’s answer changed four times, and each format drew on a different signal. We ask a consistent question — which agency should a Utah service business hire to get ranked on Google and recommended by ChatGPT — and record the businesses named and every source cited.

This week the answer opened with a local-business panel of twelve Utah agencies. Every one carried a Google review count and a star rating: Sebo Marketing at 57 reviews, GroFire at 113, Jake’s Web Design at 102, LionHead at 66, Keystone at 55, Tobe Agency at 41, down to Digital Maxima at 23 and ThunderBrand at 22. The lowest review count in the entire panel was 4. Our own profile has zero reviews, and we did not appear.

In other weeks the same question produced prose with no panel at all, and the shortlist came from sources we could actually influence: agencies’ own service pages, a competitor’s about page, and in one week our own Utah page, quoted closely enough that the positioning language was recognisably ours. The lesson is not that reviews always decide. It is that in the panel format they are close to the only thing that decides, and you cannot predict which format a given buyer will get.

Which signal matters more, reviews or content?

Neither dominates outright — they gate different answer formats, so the honest framing is which one is your binding constraint right now. A business with strong content and no reviews loses the panel. A business with 200 reviews and a thin, unsourced website gets listed but rarely gets quoted or explained well.

Signal	Which format it gates	Time to move	Who controls it
Review count and rating	Local business panels	Months	Your customers
Business listing accuracy	Local business panels	Weeks	You
Answer-first, sourced pages	Prose shortlists, AI Overviews	Weeks	You
Third-party roundups and directories	Prose shortlists	Months	Editors

The content side has the better-documented mechanics. Aggarwal et al. (Princeton, 2023) found that adding citations, quotations, and statistics to a page raised its visibility inside generated answers by up to 40%, and we have seen the on-page version of that work pay in ordinary search too: a single headline change on our Salt Lake City page preceded a move from position 73 to position 35 over four weeks, with the whole commercial query family moving alongside it.

What content demonstrably did not do was get us into a review-ranked panel. That is the cleanest natural experiment in our file, because the content work was already done and the panel still excluded us.

Why can’t an agency just get you into the panel?

Because the panel is assembled from your business listing and your customers’ reviews, and neither is an asset an agency can manufacture without crossing into fraud. Review-gating and incentivized reviews violate Google’s prohibited and restricted content policy for reviews, and purchased reviews are the fastest route to having a profile suppressed entirely.

What can legitimately be done is unglamorous and effective: claim and complete the profile, fix the category and service-area fields, and build a habit of asking satisfied customers at the moment the work lands. ChatGPT itself made a version of this point unprompted in one of our earlier samples, telling the buyer that no legitimate agency can simply place them into its recommendations and advising them to ask for before-and-after citation evidence instead of ranking screenshots.

We hold ourselves to the same standard, which is why the zero on our own profile appears in this article rather than being quietly omitted. It is our largest visible gap, it is not a content problem, and no amount of writing will close it.

What should you do first?

Fix whichever side is at zero, because the returns are wildly asymmetric at the bottom of each scale. Going from no reviews to a handful moves you from categorically excluded to eligible; going from 80 to 120 is a marginal gain. The same is true of content: a page with no direct answers and no sourced claims has a real ceiling, while a good page rewritten again is mostly effort.

For most local service businesses the honest sequence is listing accuracy first, then review velocity, then content structure, then third-party mentions. That ordering inverts what most agencies sell, because the first two are cheap and the last two are billable. It is still the right order, and any audit that skips straight to a content retainer without checking your listing and review position is answering a question you did not ask.

If you want the evidence version of this for your own category — which businesses the assistants name, what they cite to get there, and which of those doors you are closest to — that is what we do on a 30-minute call. No charge, no obligation, and you keep the findings whether or not you work with us. You can also run our free AI visibility audit first if you would rather see the scan before you talk to anyone.

What we’d actually do

Stop treating “AI visibility” as one number. It is at least two channels with different gatekeepers, and a strategy that pours everything into content while a profile sits at zero reviews is optimizing the channel that happens to be more fun to work on.

Sample your own category before you spend anything. Ask an assistant the question your buyers ask, twice, a week apart, and write down what format you get and who gets named. If you get a panel, your project is reviews and listings. If you get prose, read the sources it cites and go get into those. If you get both across two runs — which is what we usually see — you have two projects, and the cheaper one is almost always the one you have been ignoring. Our own nine-week log says the format rotates; plan for both rather than betting the year on the half you prefer.