40% of top-listed restaurants were never named by ChatGPT or Gemini

What we did

We picked four US cities (Austin, Denver, Nashville, San Diego) and four cuisines (Mexican, Thai, Italian, Indian): 16 groups. For each group we took the restaurants Google Maps lists first when you search for that cuisine in that city, 101 restaurants in all. Then we asked ChatGPT and Gemini, with web search switched on, six everyday questions such as “What are the best Thai restaurants in Austin, TX?” and “Which Mexican restaurant in Denver, CO do people recommend most?”. We asked each set twice, so every restaurant had 12 ChatGPT answers and 12 Gemini answers to show up in. A restaurant counted as named if its name appeared anywhere in the answer.

Finding 1: four in ten were never mentioned

40 of the 101 restaurants did not appear in a single one of their 24 answers. These were not obscure places: every one of them was on the first page of Google Maps results for its cuisine and city. A further 37 were named at least once but in fewer than half of their answers. Only 24 (24%) were named in half or more.

Groups are small (5 to 13 restaurants each), so treat the differences between cuisines and cities as a rough guide, not a ranking.

Finding 2: ChatGPT and Gemini disagree

ChatGPT named 59 of the 101 restaurants at least once; Gemini named 35. Only 33 (33%) were named by both. 26 were named only by ChatGPT and just 2 only by Gemini. Gemini answers tended to repeat the same two or three restaurants every time, while ChatGPT spread its picks wider. Being visible in one assistant says little about the other.

Finding 3: three restaurants take almost everything

Within each city and cuisine, we added up every mention of the sampled restaurants. In 14 of the 16 groups, the three most-named restaurants received 90% or more of those mentions. The exceptions were Thai food in Austin (70%) and Mexican food in San Diego (82%). For most restaurants, the realistic question is not “where do we rank?” but “are we one of the three?”.

Finding 4: reviews help, but do not decide it

Restaurants that were never named had a median of 545 Google reviews. Those named at least once had a median of 2,213, and those named in half or more of their answers had 2,804. So review volume matters. But it is not enough on its own: of the 28 restaurants with more than 3,000 Google reviews, 6 (21%) were never named and 11 (39%) were named in no more than one of 24 answers. Of the 17 restaurants with fewer than 300 reviews, 12 (71%) were never named.

Finding 5: the answers keep changing

Every question was asked twice. Of the 61 restaurants named at least once, 18 (30%) appeared in only one of the two rounds. A single check in ChatGPT tells a restaurant owner very little; you need repeated checks over time.

Where the AI answers came from

Both assistants searched the web before answering, and their answers named many restaurants that were not in Google’s top results at all. Citations we saw included review and listing sites such as Tripadvisor, Resy and Google Maps, and local “best of” lists. That points to what a restaurant can work on.

What restaurant owners can do

• Get onto the lists AI reads. Local food guides, “best Thai in Austin” round-ups, Tripadvisor and reservation sites are what the assistants searched. Being on them matters more than one more review.
• Say clearly what you are. Put your cuisine, neighborhood and signature dishes in plain words on your website and Google Business Profile, so an answer can match you to the question.
• Keep your name the same everywhere. Several restaurants appear under shorter or different names across the web. One consistent name makes you easier to cite.
• Earn reviews that mention dishes. Reviews help, and reviews that name what people ate give an assistant something to quote.
• Check more than once. Answers change from day to day. Our free AI visibility checker asks ChatGPT and Gemini the same kind of questions for your restaurant.

Limits

This is a snapshot taken on November 14, 2026, with gpt-5.4-mini and gemini-3.6-flash, both with web search. The sample is restaurants that Google already lists at the top, so an average restaurant is probably less visible than this. Groups are small. A mention is not always a recommendation. We tested two assistants, in English, with general questions; a diner asking about a specific dish or neighborhood may get different answers. We do not publish the names of restaurants that were not mentioned.

Cite this study

Frisler Digital, “Restaurant AI Visibility Study 2026”, November 2026, https://frislerdigital.com/statistics/restaurant-ai-visibility-study/. You are welcome to quote the figures with a link to this page. Journalists can contact us for the city-by-city breakdown.