Every week we watch AI assistants pick winners. A buyer types “who is the best mortgage broker near me” into ChatGPT, and out comes a short list of two or three names. Behind that tidy answer is a quiet ranking the model never shows you. One of the questions we get asked most by professionals trying to crack that list is simple: do reviews affect AI recommendations, or is that just SEO folklore carried over from the Google era?
To answer it with more than a hunch, we leaned on the data we already generate. Across the audits we run for service professionals and the visibility tracking we maintain for clients, we compared who the major answer engines actually named against the review profiles of those businesses and their local competitors. The pattern was consistent enough to state plainly: reviews are one of the strongest signals AI uses to decide who to recommend. They are not the only signal, and they rarely work alone, but ignore them and you are handing the recommendation to someone else.
What we looked at when we asked whether reviews affect AI recommendations
This is our own analysis, not a third-party academic study, so we want to be clear about what it is. Over several months we logged how ChatGPT, Perplexity, and Google’s AI surfaces responded to buyer-style prompts across professional-services categories like mortgage, real estate, legal, and accounting. For each prompt we recorded which businesses were named, then pulled the public review data for those businesses and a set of nearby competitors who were not named.
We were not trying to isolate reviews in a lab. Real recommendations are messy, and that is the point. What we wanted to know was whether review strength reliably separated the businesses AI mentioned from the ones it skipped. It did, far more often than chance would explain.
The signals that moved the needle
Three review attributes showed up again and again on the businesses AI named:
- Volume relative to the local market. The recommended business almost always had more reviews than the median competitor in its area, not more than every competitor everywhere.
- A strong, current average rating. Recommended businesses clustered at the top of their local rating range, and a recent dip in average tended to coincide with dropping out of the named set.
- Recency and steadiness. A trickle of fresh reviews over the last few months beat a large pile that had gone quiet a year ago.
How much do reviews actually matter versus other signals?
Reviews are powerful, but they sit inside a stack. In the cases we examined, reviews behaved less like a master switch and more like a multiplier on everything else you have in place. A business with great reviews but a messy, unclaimed profile and no quotable content still got skipped. A business with a clean profile, consistent details, and good answer-first pages plus strong reviews tended to win.
That mirrors what we explain in our overview of what answer engine optimization is and how it works: AI assembles a picture of you from many sources and recommends the business it can most confidently describe and trust. Reviews feed the trust part of that equation harder than almost anything else, which is exactly why they move recommendations.
Here is roughly how the signals stacked up in what we observed:
| Signal | Observed influence on being recommended |
|---|---|
| Review volume vs. local competitors | High — consistently separated named from unnamed |
| Average rating (current) | High — named businesses clustered near the local top |
| Review recency | Medium-high — steady recent flow outperformed stale volume |
| Owner responses to reviews | Medium — reinforced trust and freshness signals |
| Reviews across multiple platforms | Medium — corroboration mattered more than one big number |
Why answer engines lean on reviews so heavily
It helps to remember what an AI assistant is doing when it recommends a business. It cannot meet you, tour your office, or vet your work. It is pattern-matching across the public record to predict who a real person would be glad to be sent to. Reviews are the densest, most legible proof of that in the entire web. They are written by customers, attached to a verified location, dated, and full of the exact language buyers use.
That makes reviews easy for a model to read and hard to fake at scale. When ChatGPT or Gemini explains why it named you, it frequently paraphrases your reviews back almost verbatim: “clients praise their responsiveness” or “known for clear communication.” The model is not inventing that. It is summarizing the corpus you and your customers built. If that corpus is thin, dated, or negative, the model either skips you or hedges the recommendation, and a hedged recommendation rarely earns the call.
Google reviews still anchor the picture
Because answer engines lean so heavily on Google Business Profile and Maps data, Google reviews carried the most weight in what we saw. But reviews on industry platforms and aggregators were not wasted. They mattered most when they corroborated the same rating and themes, giving the model independent confirmation. A spread of consistent reviews across trusted sites proved more durable than a tall pile on a single platform.
What this looks like when a business gets it right
The publicly documented example we point clients to is Keith Akada, a Seattle mortgage broker who went from effectively invisible in AI search to the number-one AI-recommended broker in his market. Inside six weeks that shift produced roughly 30 leads and four closed deals. Reviews were not the only thing we worked on, but they were a load-bearing part of it, because they gave the assistants concrete, current, human proof to point at when buyers asked who to trust.
The takeaway is not “collect reviews and wait.” It is that reviews convert into recommendations fastest when the rest of your foundation is solid, so the model has somewhere clean to attach all that trust. If you want the wider playbook around this study, our state of AI search for professional services report covers how these signals are trending across categories, and our look at who ChatGPT recommends when buyers ask for a mortgage broker shows the review pattern playing out in one vertical.
How to turn this into action
If reviews move AI recommendations, the practical question is what to do Monday morning. Based on what consistently correlated with being named, here is the order we recommend:
- Claim and clean your Google Business Profile. Reviews can only help if they attach to a complete, accurate, claimed profile. This is the foundation everything else multiplies against.
- Build a steady review habit, not a one-time push. Ask every satisfied client, make it frictionless, and aim for a consistent monthly flow so recency never lapses.
- Protect your current average. Address service issues before they become reviews, and respond to the reviews you do get, especially the critical ones.
- Spread reviews across the platforms buyers and AI trust so your reputation is corroborated rather than concentrated.
- Pair reviews with answer-first content and consistent business details so the model has clean, quotable material to attach the trust to.
None of these are exotic. That is the encouraging part. Reviews are one of the few AI-visibility levers a busy professional can start pulling today without rebuilding a website, and one of the few where the work you do also genuinely improves the business behind it.
So, do reviews affect AI recommendations?
Based on the data we collect, the answer is a confident yes, with an asterisk. Reviews are among the strongest signals deciding who AI names, but they are a multiplier, not a magic word. They reward businesses that are already trustworthy and well-documented, and they punish quietly by simply leaving the under-reviewed off the list. The good news is that this is a winnable game: it rewards steady, honest effort over budget, and the businesses showing up in AI today are usually the ones that started treating reviews as a visibility asset rather than an afterthought.