Most of the advice in this field asks for things a five-person firm cannot get: a Wikipedia entry, a Reddit following, press coverage. A review profile is the rare lever that is both cheap and measured. The measurement comes with two qualifications, and both are worth reading before you act on it.
In May 2026 Seer Interactive published the largest study so far on reviews and AI answers. It ran 15,783 prompts through ChatGPT, Google AI Mode, Gemini and Perplexity, collected 804,491 responses across 1,926 brands in eight industries, and sorted the brands by how much of a Trustpilot profile they had: none, minimal, active, or actively managed (Seer Interactive). The gap between the first two groups is the finding everyone quotes.
The shape of that curve matters more than the headline. Almost all of the movement happens between zero reviews and a dozen. After that, volume still helps, but at a fraction of the rate. And the effect held for the brands least able to buy their way in: a brand with a Domain Rating of 0 to 15 and a minimal profile reached a 54.5% citation rate, against 52.6% for a mid-authority brand at the same tier (Seer Interactive). A profile did for small brands roughly what years of link building did for larger ones.
Seer also traced how those citations arrived. When ChatGPT answers a question about a company, it runs several background searches first, and "reviews" appeared in 60.1% of those searches. Only 0.5% of Trustpilot citations came from ChatGPT deliberately searching Trustpilot by name. The rest arrived because a Trustpilot page happened to rank for whatever the model searched, or because the model already knew the profile existed (Seer Interactive). That is the mechanism: the review platform's own search visibility carries you. You do not need hundreds of reviews to be found. You need a page on a domain that ranks whenever someone types your name plus "reviews".
The absence has a cost of its own. Seer found ChatGPT stating it outright in answers, with wording along the lines of a brand lacking reviews on trusted sites, and in several cases citing the brand's own website in the same answer while flagging that no independent reviews existed (Seer Interactive). The name was visible. The credibility was not.
First, the study was published with Trustpilot, and Trustpilot issued the press release (Seer Interactive). That does not make the data wrong, and Seer describes its method in enough detail to check. It does mean the study measured one platform and nobody has yet published the same test for Google reviews, ProvenExpert or a trade-specific portal. Second, the metric is narrower than "cited by AI". It counts how often a brand appeared alongside a Trustpilot citation. Read it as evidence that the profile becomes part of the answer, not as a 53% chance of being recommended.
Seer's own denominator also disagrees with a larger dataset. Yext categorised 6.8 million citations from 1.6 million AI responses and found reviews and social content at 8% of the total, rising to 13.3% in food service, while brand websites took 44% and listings 42% (Yext). Seer's aggregate for review and trust sites was 14%, but that hides a range: 1.5% for awareness-stage questions and 24.3% for questions with buying intent (Seer Interactive). Reviews are close to irrelevant when someone asks what a service is, and the second-largest source when they ask whether to buy it from you.
SOCi's 2026 Local Visibility Index looked at close to 350,000 locations across 2,751 multi-location brands. Those brands appeared in Google's local three-pack 35.9% of the time. ChatGPT recommended 1.2% of the locations, Perplexity 7.4%, Gemini 11% (Search Engine Land). The locations that did get recommended were not random. On ChatGPT they averaged 4.3 stars, on Perplexity 4.1, on Gemini 3.9. Financial brands with ratings around 3.4 and review response rates below 5% were, in SOCi's phrase, effectively invisible.
So the two studies describe two different jobs. A profile gets you into the pool of businesses the model considers. A rating decides whether you leave it. SOCi's sample is chains rather than independents, and the thresholds will differ by category, but the direction is consistent with Seer's finding that ChatGPT read the TrustScore out loud in 59% of the answers where it cited Trustpilot at all (Seer Interactive).
Claim a profile on whichever review platform ranks for your name plus "reviews" in your market. In Germany that is often Google first, then a trade portal, then Trustpilot or ProvenExpert; check the actual search result rather than guessing. A free profile with a dozen real reviews is the cheapest measured gain in this field. Ask every satisfied customer, and reply to every review, including the bad ones, because response rate is one of the fundamentals SOCi tied to visibility.
Do not buy reviews, seed them from staff, or offer a discount for five stars. In the United States the FTC's rule on consumer reviews bans all three, including reviews written by people who never used the service and any incentive tied to a particular sentiment (FTC); German competition law reaches the same practices from a different direction. The models also paraphrase the reviews they find, and Seer saw negative themes, billing disputes and unmet promises among them, surface in answers just as readily as praise. A profile amplifies what is already there.
And keep the number in proportion. A review profile is an entry ticket for questions that carry buying intent. It does nothing for the earlier question, the one where someone asks the model what kind of business they need at all. That answer still comes from your own pages, as it has in every study we have covered here.
We ask ChatGPT, Perplexity, Google AI and Claude the questions your customers actually ask, and send you the score. No account, no card.
run_check →