Searchable asked ChatGPT, Google Gemini, Google AI Mode and Perplexity where to eat in 55 US cities, collecting 174,142 answers to general, diner-type and cuisine questions. We measured how often AI recommends the same restaurants, which ones it names most, and how many of each city's highest-rated restaurants it leaves out.
1 in 4
frequent US AI chatbot users ask AI to find them a place to eat
1 in 12
highly rated restaurants are invisible to local AI recommendations
57%
of AI restaurant recommendations repeat the same five restaurants
- (01)The cities where AI recommends the same restaurants most often
- (02)How often asking again gets you a restaurant AI hasn't already named
- (03)The restaurants AI recommends most in every city
- (04)How many of each city's highest-rated restaurants AI recommends, and how many it never names
- (05)How often AI repeats itself when diners ask for a cuisine
- (06)How the four AI chatbots differ in what they recommend
- (07)Which websites AI cites when it recommends a restaurant
How often do the same five restaurants get recommended in US cities?
The more often AI repeats the same few names, the harder it is for any other restaurant to get into the answer. Across all 55 cities, a city's five most-recommended restaurants appear in 56.6% of answers to the everyday questions: the best restaurants, where to eat, and where to eat on a first visit.
Tampa is the most repetitive: its five most-recommended restaurants appear in 70.9% of answers, followed by Santa Fe and Key West. San Diego is the least, and even there each of its top five turns up in two in five answers. Hover over a city to see its figures.
- (01)Tampa70.9%
- (02)Santa Fe70.2%
- (03)Key West69.5%
- (04)Savannah67.4%
- (05)Anchorage66.6%
- (01)San Diego41.1%
- (02)New York City44.4%
- (03)Dallas46.4%
- (04)Houston46.6%
- (05)St. Louis47.5%
The cities where AI chatbots recommend a new restaurant most and least often
Across the 55 cities, a third of repeat answers (32.5%) named a restaurant that no earlier answer had. Ask again in Anchorage and only 21.3% of answers name a restaurant AI hadn't already suggested. There, and in Santa Fe, Branson and Key West, asking again mostly brings back the same list. New York City, Las Vegas and San Francisco give the most variety.
Share of repeat answers to the three main “where to eat” questions that name at least one restaurant no earlier answer to the same question had named, with answers taken in the order they were collected and each chatbot weighted equally. Only real restaurants count: every name that was new in an answer was checked, and headings, dishes and invented places removed.
The 10 most recommended restaurants in each US city
In many cities one restaurant comes close to a guaranteed mention. Tampa's Columbia Restaurant is named in 85.8% of AI answers about where to eat there, and Savannah's Olde Pink House and Indianapolis's St. Elmo Steak House in more than 4 in 5. Choose a city to see its ten most-recommended restaurants, including the ones AI names that are not on Tripadvisor's list.
See the 10 restaurants AI recommends most in every city
Enter your details to open the ranking for all 55 cities, including the restaurants AI names that are not on Tripadvisor's list.
No spam, ever. We respect your privacy.
How many of each city's highest-rated restaurants AI actually recommends
Pick a city and drag the handle across the map. Left of it, each of the city's top-ranked restaurants takes its colour from how often AI names it. Right of it, every restaurant is shown the same, as diners rank the city. Across the 55 cities, AI regularly recommends 1 in 8 of these restaurants, and 1 in 12 never comes up in a general answer. In New York City it regularly recommends 5%: the city's highest-rated restaurants on Tripadvisor lean toward tourist favorites such as Bubba Gump Shrimp Co. (#3), while AI's own New York favorites, such as Le Bernardin (#105), sit outside them.
Ask AI for a steakhouse and the same five come back in 73.5% of answers
We also asked each chatbot for the best restaurants of nine kinds of food in every city. Narrowing the question does not widen the answer: for every cuisine, a city's five most-named restaurants for that cuisine appear in more than half of the answers to it. Steakhouses, Italian and seafood are the most repetitive. Choose a city to see its own ranking.
Average share of answers to each cuisine question naming that question's five most-recommended restaurants.
ChatGPT names about twice as many restaurants per answer as Google AI Mode
Ask where to eat and ChatGPT names about 11 restaurants in an average answer, Gemini 6, Google AI Mode 5 and Perplexity 2. That counts every restaurant named, on Tripadvisor's lists or not. The chatbots also disagree on which ones: for the same question about the same city, their top 10 lists overlap by only 30%.
Average number of restaurants named in each answer to the three main “where to eat” questions, on Tripadvisor's lists or not. Headings, dishes and neighbourhoods are discounted using a hand check of 100 answers per chatbot.
Eater and Tripadvisor are the websites AI cites most when it recommends a restaurant
Each chatbot shows the websites it drew on. Eater (34.8%) and Tripadvisor (34.5%) appear in the most answers. Google AI Mode leans on Google's own business listings more than on any outside site, and Reddit appears in more than 1 in 8 answers. Our guide to AI search optimization for local businesses covers the listings and profiles these answers draw on. Choose a city to see which local guides AI relies on there.
Share of answers citing each website at least once, per chatbot and then averaged across ChatGPT, Google AI Mode and Perplexity. Google place listings are Google's own business profiles, linked from AI Mode answers. Gemini is not included: only 1 in 10 of its answers showed sources.
Where restaurants can start showing up in AI answers
AI is already part of how Americans choose where to eat. One in five US consumers use AI tools such as ChatGPT to find restaurants and bars, rising to 45% of 35 to 44 year-olds, according to a survey of 763 US diners by Reputation and Nielsen (October 2025).
The findings above describe AI's recall problem: a restaurant that is not in a chatbot's small pool of names does not get considered when it assembles an answer. Here are five things that can help with that.
Ask the chatbots your own diners' questions and write down who gets named
Ask the plain version first, then ask again as familiar diner profiles, such as a couple booking an anniversary dinner or a family with young children. Note which restaurants are named ahead of you, and treat them as your competitors for AI visibility.
Build your Google reviews and your rating
AI chatbots favor restaurants with stronger reviews. We matched 4,744 of each city's highest-rated restaurants to their Google listings, and those rated 4.7 stars or higher are named more than four times as often (8.2% of answers) as those rated under 4.5 (1.9%).
Review count matters too. Restaurants with more than 4,000 Google reviews are named over five times as often (7.0%) as those with fewer than 500 (1.2%).
Both hold when the other is held level, in almost every city and for all four chatbots. Google AI Mode, which draws on Google's own business listings, is the most sensitive to both, while ChatGPT pays the least attention to review count.
This shows which restaurants AI favors rather than proving that reviews earn the mention, but the pattern is consistent. Ask happy diners for a Google review, and reply to the ones you get.
Answer what diners actually want to know, and say it the same way everywhere
Do you take reservations? What are you known for? A specific answer repeated everywhere (“wood-fired Neapolitan pizza, walk-ins only, open until midnight”) is something a chatbot can match to a question.
Keep it current on your website and every listing diners see, including your Google Business Profile. Google says businesses with complete and accurate information are more likely to show up in local search results.
Publish your menu as text on your website, not only as a PDF or photo, so search engines and AI tools can read it dish by dish.
Give AI something to match for long, specific questions
Google's AI Mode and AI Overviews split one question into several related searches, which Google calls query fan-out. ChatGPT similarly rewrites a prompt into one or more targeted searches.
So “a quiet date spot with vegan options and good cocktails” becomes a search for each of those things. Describe your dishes and the occasions you suit on your own website, so there's something to match.
Get the sources AI reads describing you
The websites in the chart above are where AI's descriptions come from, with Eater cited in 34.8% of answers and Tripadvisor in 34.5%. Reviews, local food guides and press coverage are where your restaurant gets repeated back.
None of this guarantees a mention. Closing the gap between being a great place to eat and being in the answer is the work of AI visibility, also called answer engine optimization (AEO). It's what Searchable measures for more than 1,000 brands every day.
See how often AI recommends your brand
Track your visibility across ChatGPT, Gemini, Google AI Mode and Perplexity.
How we did it
Searchable asked ChatGPT, Google Gemini, Google AI Mode and Perplexity for restaurant recommendations in 55 major US cities and tourist destinations over six days in October 2026. This produced 174,142 answers on the best places to eat in general, for nine types of cuisine and for other requirements such as hidden gems and independent restaurants.
For each city, we measured how often AI repeats the same restaurants, and what share of the city's highest-rated restaurants it recommends regularly, sometimes or never.
A city's highest-rated restaurants are those at the top of Tripadvisor's ranking for it, rated 4 stars or more with at least 50 reviews. We checked each one on Google Maps and removed any that had closed, were not restaurants or sat outside the city, refilling from Tripadvisor's next-ranked restaurants where its list allowed. That left between 52 and 98 restaurants per city. Because the lists differ in length, one city shouldn't be compared with another on this measure.
To size demand, we classified every prompt that 75,308 US users of ChatGPT, Gemini, Google AI Mode and Claude sent between 1 January and 20 September 2026 in Searchable's opt-in consumer panel. Among frequent users, the 6,638 who sent 250 or more prompts over the period (roughly one a day), 1 in 4 (25.6%) asked AI to recommend or find somewhere to eat.
