How a local business gets recommended by ChatGPT and where Google reviews come in
A local business doesn't need separate "AI optimization" to show up in ChatGPT recommendations. The assistant leans on the same data as Google Maps - the Google profile with its rating and review count - plus third-party directories and lists. What matters is fresh, specific reviews, an accurate category and area in the profile, and the same details everywhere. Below is our own check from November 6, 2026 of whom ChatGPT recommends
Where do ChatGPT and Gemini get information about a salon or a garage?
ChatGPT doesn't disclose how it picks businesses, but you can see what it shows. In our check on November 6, 2026, every answer about a dentist, a car repair shop or a beauty salon included a rating and review count in the Google Maps format, for example "4.9 · 2,941", with a note that the ratings came from map or search data at the time of the search. When asked where the numbers came from, ChatGPT first warned not to treat them as verified Google Maps ratings, and after checking again named Google business listings and a few local directories - where the counts lagged behind Google: 486 reviews for one salon in a directory against 511 in its profile
Outside write-ups describe the same picture. Reviewly's playbook from November 4, 2026 calls the Google Business Profile the primary data source: Google AI Mode and Gemini read it directly, and ChatGPT uses live web search plus licensed location data, with Foursquare supplying place data. The same piece calls reviews the richest source of third-party language about a business. GatherUp, on November 3, 2026, summarized research by Andrew Shotland of Local SEO Guide and the rank tracker Nozzle: YouTube came out as the second most cited domain in local AI answers
We didn't test Gemini: it requires a Google account login, and we wanted a clean run with no search history. Everything about Gemini here is Reviewly's observation. The rest of the article is about ChatGPT
Whom did ChatGPT recommend in our check?
Method: chatgpt.com without logging in, free mode and the default model, a new chat for every run, desktop Chrome, one city (Kyiv, Ukraine, where we are based), November 6, 2026. Three prompts - a good dentist, a car repair shop, a beauty salon - two runs each, six runs in total. Ratings and counts below are exactly as ChatGPT showed them. Business names are a snapshot, not our recommendation
| Category | Named by ChatGPT (rating · reviews) | Repeated in both runs |
|---|---|---|
| Dentist | EUROZET (4.9 · 2,941), Klub 32 (4.9 · 1,253), SANOS (5.0 · 671), Healthydent (4.8 · 792), Lukashuk Dental Clinic (4.9 · 2,217) | 5 of 5, nearly the same order |
| Car repair | Kurator.UA (4.9 · 2,443), Bosch Auto Service Avtopan (4.8 · 2,223), Oiler Pechersk (4.8 · 1,881), Elcars (4.8 · 975), Stop&Go (4.8 · 339) | 2 (Kurator.UA and Bosch Avtopan); the runs named 4 and 3 shops |
| Beauty salon | Salon Special (5.0 · 511), Cherry Beauty Studio (4.9 · 605), BACKSTAGE Lobanovskoho (4.8 · 642), Sweet Beauty Family (4.9 · 402), DVI Beauty Space (4.8 · 288), Maija Lux (4.9 · 190), Andri beauty space (5.0 · 110) | 3 of 5 (Salon Special, Cherry Beauty Studio, Andri) |
The same day we opened three of these profiles in Google Maps by hand
| Business | In the ChatGPT answer | In Google Maps, November 6, 2026 |
|---|---|---|
| Salon Special | 5.0 · 511 | 5.0 · 511 |
| EUROZET | 4.9 · 2,941 | 4.9 · 2,936 |
| Kurator.UA | 4.9 · 2,443 | 4.9 · 2,444 |
So the numbers in ChatGPT's answers were the Google profile's rating and review count, a few reviews behind. The limits of the check are plain: one day, one browser without an account, one city, three categories, two runs each. It's a snapshot, not a study, and in your city and a month from now it may look different
How many reviews did the recommended businesses have?
Across all six runs, every business named had a rating between 4.8 and 5.0 and between 110 and 2,941 reviews. None had a dozen reviews. None had 4.3 either, so the popular claim that "4.3 with a hundred reviews beats 5.0 with ten" was neither confirmed nor refuted by our check
ChatGPT explained its picks with good ratings and a noticeable number of reviews, and offered to narrow the search by area and service - so category and location in the profile matter here too. In the salon chat it also suggested comparing places by Google Maps rating, number of real reviews and negative reviews from the last 3-6 months
NiceJob, writing for home service businesses, gives another benchmark (the page has no publication date): a rating of roughly 4.2-4.7 with no hard line, and 4 to 8 new reviews a month. Citing a 2026 Trustmary study, it says businesses recommended by ChatGPT average 4.3 stars. In a crowded category in a big city, treat those numbers as a floor rather than a goal. The practical benchmark is the competitors already showing up in your own results
What can you do yourself in a month?
- Check the category and address in your Google profile. In our runs ChatGPT asked for the area and service and showed addresses in its answers. Your primary category should name what people actually ask for: "dental clinic", not "medical center". Address, phone and hours should match your website and directories, so the assistant has nothing to mix up
- Collect fresh, specific reviews every week, not once a quarter. Reviewly calls review text the main third-party language a model uses to match a business to a request, so "cleaning took 40 minutes, no pain" gives it more than "all great". Ask right after the visit with a link that opens the Google form even from a messenger's built-in browser. With Pochuto, a low rating first opens a form for the owner, so the problem reaches you before it reaches the profile, and the Google link stays available. Don't dictate what to write: Google's policy lists merchants asking staff to solicit reviews with specific content as rating manipulation
- Reply to reviews, negative ones first. In the salon chat ChatGPT told people to look at negative reviews from the last 3-6 months. A specific reply from the business about what you fixed stays next to the review the assistant tells people to read
- Be listed in the directories and lists the assistant cites. Ask ChatGPT where its ratings come from for your category, then check that you are in those sources with the same name, address and phone as on Google. A YouTube video with your business name and area is a mention too: in the research GatherUp summarized, YouTube was the second most cited domain in local AI answers
We don't promise a timeline. We don't know how often ChatGPT refreshes its data, and it doesn't disclose how it chooses. A month is enough to tidy up the profile and start a steady flow of reviews; whether you appear in answers is something to check again with the same prompts
See how the review page worksWhat should you avoid?
- Buying reviews. Google's policy treats reviews paid for, directly or in kind, as fake engagement, and since ChatGPT takes the rating and count from Google, a removed review doesn't count in its answers either
- Writing reviews for customers with ChatGPT. Google's policy bans content not based on a real experience, and a batch of similar texts from one day looks exactly like that
- Building separate "for AI" pages and a text for every query. The sources in our runs were Google profiles, directories and lists, not the businesses' own sites. Reviewly says the same: none of these engines reads your homepage to form an opinion - they build the picture from other people's pages
Sources
- Reviewly, November 4, 2026: LLM SEO for local businesses in ChatGPT, Gemini and Perplexity
- NiceJob, undated: why a perfect 5-star rating might hurt your AI search visibility
- GatherUp, November 3, 2026: reviews, AI search and YouTube
- Google Maps user contributed content policy: fake engagement and rating manipulation
A free first step
See how your profile looks to a customer right now
Your rating, review count and how many five-star reviews it takes to reach the next tenth - in 30 seconds, no sign-up