Do Google reviews matter when AI picks a business?
In our test of 733 AI answers, the businesses named most had a median of 220 Google reviews, the others 82. What that shows, and how to earn reviews honestly.
Google reviews line up with who AI assistants recommend, but they don't decide it on their own. In our test of 733 answers from ChatGPT, Gemini, Perplexity and Claude across 11 Australian areas, the three businesses named most in each area had a median of 220 Google reviews, while the other businesses the assistants named had a median of 82. Reviews are one of the few signals you can see, compare and improve yourself, so a steady, honest flow of them is worth the effort.
What the numbers show
For every area we tested, we took all the businesses the four assistants named and sorted them by how often they were named. The three at the top of each area, 33 businesses in all, had a median of 220 Google reviews. The other 130 businesses that got named at least once had a median of 82. That's nearly three times as many reviews for the businesses that were named most.
The top spot matters more than it sounds. In a typical area, the business named most appeared in about half of all the answers for that area. So the gap between being named now and then and being named in half the answers is real, and the businesses on the right side of it tend to have more reviews.
Notice the other number too. The businesses named less often still had a median of 82 reviews. Being in the answer at all did not need hundreds; being named again and again went with having more.
What the numbers don't show
They don't show that reviews cause the recommendation. A business that has been busy for ten years tends to have more reviews, more pages that mention it, an older website and more of everything else an assistant reads. Reviews travel with all of that, and our data can't pull them apart.
We also only counted businesses that were named. A business the assistants never mentioned isn't in these figures, so we can't tell you what happens to one with five reviews or none. And we counted reviews, not star ratings or how recent they were.
Eleven areas and three trades (dentists, physiotherapists and hair salons) is a small sample. Read it as a first look, not a rule.
Where your reviews sit in what AI reads
Your reviews live on your Google business page, and that page is one of the two things the assistants read most. Of the 4,657 sources they cited in our test, 60% were businesses' own websites and 18% were Google Maps and Google business pages. Directories came next at 8%.
So when an assistant looks you up, there is a good chance it lands on the page that shows your review count, your rating and your replies. Its details need to match your website too; we wrote about that in your Google business page is what AI reads.
How to earn reviews honestly, this month
These steps stay inside Google's rules, which matter: Google's review policy bans offering payment, discounts or free goods for reviews, and bans asking only your happy customers.
- Get your review link. In your Google business profile, use the option to ask for reviews; it gives you a short link that opens the review box directly.
- Ask while the visit is fresh. Send a text or email the same day, with the link and one plain line: "If we looked after you well, a short Google review helps other people find us."
- Put the link where people already are: a QR code at the counter, on the receipt or appointment reminder, and in your email signature.
- Ask everyone, not only the people you expect to be happy. Picking only the happy ones is against Google's policy.
- Never pay for reviews or trade a discount for one. Google can remove reviews that break its policy.
- Reply to every review, good and bad, calmly and in a sentence or two. A reply is part of the page the assistants read, and part of what the next customer reads too.
- Check where you stand. Ask ChatGPT "best [your service] in [your suburb]", then look up the businesses it names on Google Maps and note their review counts. That tells you the gap in your own suburb, which matters more than our median.
A few new reviews every month beats a burst once a year.
What we're doing on sbw.studio
We are at the other end of this. sbw studio has no Google reviews and no Google business page yet, and when we asked the four assistants eight questions about the work we do, none of the 32 answers named us.
For a studio like ours, reviews may not be the first thing to fix. When we asked Perplexity two of our questions, it cited agencies' own pages and lists of agencies written by other agencies, not review pages. So we're putting our time into pages that answer the questions people ask, like this one, and into getting other sites to mention us. We have still set ourselves five real reviews by the end of November, and we'll say in our monthly growth log whether we got there and whether anything moved.
A local clinic or salon is in a different position, and it's the one our data covers. For those businesses, the ones AI named most had far more reviews.
How we measured
On 5 October 2026 we ran our AI Recommendation Report method in 11 areas across five states, around 12 local businesses: five dentists, four physiotherapists and three hair salons. We asked each assistant, with web search on and its location set to the suburb, the questions a customer would ask, then took each named business's Google review count from its Google Maps listing. The full findings and method are at score.sbw.studio/findings.
For why only a few businesses get named at all, read ChatGPT names about 3 businesses.
If you'd like to see where you stand, the AI Recommendation Report asks the four assistants about your business and your suburb, shows who they name instead of you, and tells you what to fix first.