The short answer: they go together, but nobody has proven cause
If you are asking whether reviews affect AI recommendations, the honest answer has two parts. Studies find that businesses AI assistants name tend to have many more reviews and better ratings than those they skip. And no study we found shows that adding reviews makes an assistant name you, or says how an assistant weighs them.
This page lists what each source measured, where it falls short, and what it means for a service business. It is part of our guide to how customers see you, which covers what customers find when they look you up.
What the studies found
| Source | What it measured | Finding | Limits |
|---|---|---|---|
| Insites AI Visibility Report (August 2026) | 10,000 US local businesses on ChatGPT and Perplexity | Businesses surfaced by both averaged 133.4 Google reviews, against 10.7 for those surfaced by neither. Review count was the strongest correlating factor. | Correlation only. Insites sells AI visibility reports, and some of its figures are inconsistent. |
| Insites, by tier | Reviews found across the web, by how often a business was recommended | Highly recommended businesses averaged 304.2 reviews, against 117.5 for those never recommended. | The article calls the gap “near 3x”, but the numbers work out to about 2.6x. |
| Insites, ratings | Average Google rating | 3.8 for the top three tiers, 3.3 for never recommended. Rating was a comparatively weak predictor. | Small gap. The article also gives 1.7 stars for businesses found by neither platform and does not reconcile it with the 3.3. |
| SOCi Local Visibility Index (January 2026) | About 350,000 locations of 2,751 multi-location brands | ChatGPT recommended 1.2% of locations, and the locations it recommended averaged 4.3 stars. SOCi says ratings and reviews are among the strongest influences. | Large multi-location brands, not single-location shops. |
| Yext citation study (July to August 2025 data) | 6.8 million citations from ChatGPT, Gemini and Perplexity | Reviews and social made up 8% of citations. First-party websites were 44% and listings 42%. In food service, reviews and social were 13.3%. | Four industries only (retail, financial services, healthcare, food service). Yext sells listings management. |
How to read the table together: review count and a decent rating travel with AI visibility. But in the Yext data, review pages are a small share of what assistants cite directly. The websites and listings that carry the facts matter at least as much.
What the evidence does not show
- Cause. A business with 133 reviews is usually also older, busier, better listed and easier to find. Reviews may be one signal among many, or just a sign of a healthy business.
- A threshold. You will see claims online that you need 30, 100 or 150 reviews. We could not open any source that supports a fixed number. The Insites averages describe groups of businesses. They are not a cut-off.
- Review wording. We found no study that tests whether the words in reviews change AI recommendations. Anyone who says they know is guessing.
- How each assistant weighs reviews. SOCi notes that Gemini relies solely on Google Maps for location information, while ChatGPT and Perplexity pull from a wider set of sources. BrightLocal’s citation data shows Google Business Profile and Yelp cited most. Perplexity was reported in March 2024 to have added Yelp data to its local answers. Beyond that, the weights are not public.
Google’s own documentation helps with the Google side. Google says that for local results, prominence draws on how many reviews you have, and that more reviews and positive ratings can help your local ranking. Google’s guide for AI features says Business Profiles can help businesses appear in AI responses, and does not discuss reviews. So the link from reviews to Google’s AI answers is plausible but not stated.
Why reviews pay off even if AI ignored them
People check. BrightLocal’s 2026 Local Consumer Review Survey of 1,002 US adults found:
- 97% read reviews for local businesses, and 41% always do.
- 45% had used ChatGPT or other generative AI for local recommendations, up from 6%. 40% trust AI platforms for business recommendations and 32% do not.
- 82% read AI-generated review summaries, and 23% would rely on the summary alone.
- 68% require four stars or more, and 47% will not use a business with fewer than 20 reviews.
- 74% only look at reviews from the last three months, and 56% named “backed up by other reviews” as the top factor in trusting a review.
- 80% are more likely to use a business that responds to all reviews. 50% are put off by templated replies.
So the sequence is often: the assistant names you, then the customer checks your reviews before calling. A thin or stale review profile loses the customer at that step, whatever the assistant did.
What to do about reviews
These steps follow from the data above, plus our own judgment.
- Count where you stand. Write down your Google review count, average rating, newest review date and the same numbers for two competitors. If the gap is 10 against 130, that is your priority.
- Ask at the right moment. Ask right after a good job, by text or in person, with a direct link. Spread requests over time. Recent reviews matter to customers.
- Reply to every review, in your own words. Name the service. Thank people. Handle bad reviews calmly and offer to fix the problem.
- Cover more than one site. Studies show assistants citing Google, Yelp and Facebook. Keep each profile claimed and accurate. See Apple Business Connect, Bing Places and Yelp.
- Keep the facts consistent. SOCi found hours and address errors in ChatGPT and Perplexity answers. Reviews help little if the profile lists the wrong phone number. Use the Google Business Profile checklist.
- Do not buy or invent reviews. It risks your profile and your name, and customers can tell.
Review text is worth a thought, even without a study. Customers say they trust reviews that agree with each other, so asking people to describe what you did for them, in their own words, gives other customers (and AI summaries) real detail to read. That is our judgment, not a measured effect.
A note for clinics
Dentists and med spas should be careful when replying to reviews. Do not confirm that the reviewer was a patient, and do not mention treatments, dates or health details in a public reply. A neutral reply such as “Thank you for the feedback. Please call us so we can talk it through” is safer. Rules vary, so ask your compliance adviser. This is not legal advice.
Check what AI says about you
Reviews are one of several signals. To see where you stand, run the prompts in our 30-minute self-test, and read how ChatGPT and Google AI pick a med spa, dentist or contractor for the sources each trade sees cited. If a competitor keeps appearing, what to check and fix lists seven steps.
Nobody controls what ChatGPT, Google AI Overviews, AI Mode or Perplexity recommend. We do not promise placement. We check your profile, your listings and what the assistants say today, and repeat the check later so you see a before and after. A free demo is the quickest way to start.
Published . Updated .