Which restaurants does AI recommend in the US?

What we asked

50 questions written the way US diners ask them. The 40 English questions were spread over eight cities, 5 each: Philadelphia, Nashville, Seattle, Chicago, Tampa, Austin, San Antonio and Charlotte. They cover ten kinds of request (birthday dinner, Mexican, family brunch, cheap but good, pizza, a coffee shop to work from, romantic with a view, healthy meal prep delivery, open late and steakhouse) and are tagged by intent: 20 specific, 12 compare, 4 price and 4 urgent. The 10 Spanish questions were set in Miami and New York (5 each), tagged "mixed". Six of them, word for word:

How often AI named a restaurant

Every one of the 50 answers named at least one restaurant, and none hedged. Only fitness matched that. The lists were long whatever the question was:

This is the opposite of most industries. Across the study, English price questions were the weak spot, with 17.6% hedged. For restaurants, "cheap but good" got the longest lists of all, because city media publish cheap eats guides and the search finds them. The shortest answer named two places: the Spanish question about a family breakfast in Miami, built largely from blogs and travel pages, named Denny's and one other.

Our people check found no individuals among the names. Even with long lists, Claude stayed careful: in at least 14 answers it warned that its sources were marketing or low quality, and in at least 15 it said there was no single best option.

Where the answers came from

Claude cited 453 sources for the 50 restaurant answers, 9.1 per answer, from 219 different websites, the widest spread of any industry in the study. We sorted every one into the same nine types as the main study.

Media showed up in 46 of the 50 answers and made up most of the sources in 19. No single outlet dominated: the 193 media citations came from 96 outlets, led by Hoodline (10), the San Antonio Current (9), The Infatuation, Austinot and Miami New Times (7 each). That is good news for an independent restaurant: the outlet that matters is your own city's.

The one big directory was OpenTable, with 73 citations in 18 answers. Only 6 were on opentable.com. The other 67 were on OpenTable's sites for other countries, such as Australia (14), Ireland (12), the UK (10) and Singapore (8), carrying the same US lists: "birthday restaurants Chicago", "best Mexican restaurants Austin". Several were about nearby towns rather than the city asked, such as Gastonia for Charlotte and Bradenton for Tampa.

The spam layer was heavy too: 82 sources from 32 unrelated websites, in 27 answers, and most of the sources in 6 of them. Described without naming the sites, which are often victims: a cancer charity's mobile site with "Best Pizza Places in Philadelphia", a hardware supplier's subdomain with "Top 10 Best Brunch in Chicago", a recovery group's meeting finder with a Nashville steakhouse guide, and a company app server with "Best Of 11 Birthday Dinner In Philadelphia". The same titles sat on several hosts at once. In Chicago, 35.6% of restaurant sources were pages like these.

Review sites were mentioned, not read. 8 answers told diners to check Yelp, 3 pointed to Google reviews or Maps, and none cited a page on either.

The restaurants named most often

The 50 answers named 368 different places in 384 mentions, and 96.2% were named only once. These were named most often in our answers:

The top five took 3.1% of mentions, the lowest share of the 10 industries (dentists: 15.2%). Factor, a national prepared meal delivery service rather than a restaurant, came up in three of the four meal prep questions, once through an affiliate comparison page that the answer flagged as such. Everywhere else, each answer drew its own names from whichever local guides matched the question. A place that is in the right "best brunch for kids" or "open late" guide gets named for that question, whatever its size.

City by city

Each city had 5 restaurant questions, so these are notes, not rankings. Every city got names in every answer.

• Seattle and Austin had the cleanest sources. In Austin, a local guide (Austinot) was cited 7 times in 3 answers.
• San Antonio leaned on its alt-weekly: the San Antonio Current was cited 9 times in 3 answers.
• Chicago and Charlotte had the most spam. A Chicago late-night answer warned of templates with "Insert name of a local pizza place".
• Nashville: 8 of the 9 sources for "Best steakhouse in Nashville?" were off-topic hosts.
• Philadelphia: the late-night answer leaned on The Philadelphia Inquirer and Philadelphia Magazine.

What the Spanish questions showed

Restaurants were one of the strongest industries in Spanish. All 10 Spanish questions got Spanish answers with names, and all 10 cited at least one page our check counted as Spanish: 33 of 91 sources (36.3%), above the 23.8% across all Spanish answers in the study. Spanish-language media and travel writing carried them: El Diario NY, Univision, Spanish travel blogs, and NYC Tourism's Spanish pages. One caution on our own number: 7 of those 33 pages were really Portuguese, from a Brazilian travel blog and forum. We found no Spanish answer that drifted to another country.

Miami answers pulled from local media in both languages. Taquiza was named in 3 of the 5 Miami questions, all in Spanish, and Miami New Times was cited 7 times.

Three real answers

What a restaurant should do

1. Get into your city's own guides. 96 outlets fed these answers, mostly local: city magazines, alt-weeklies, Eater, The Infatuation, Time Out, Hoodline. Tell their food writers when you open, change the menu, add brunch or extend hours. We do not buy or promise placements in them.
2. Be specific about what you are good for. The questions asked for birthdays, kids, a view, late nights and cheap eats, and the guides that answered were titled the same way. Say it plainly on your site and in your listings: private room, kids' menu, kitchen open until 1 a.m.
3. Complete your OpenTable profile, if you take bookings. Its pages fed 18 answers, and the Tampa birthday answer quoted ratings and reviews from it. Fill in cuisine, neighborhood, price range and occasion details.
4. Give your own site one page per thing guests ask. Menu with prices, hours as plain text, the neighborhood and the city and state. Own sites were only 14.6% of sources, so a clear page has little competition.
5. Use one spelling of your name everywhere. A Nashville answer noted that Capital Grille was spelled "Capitol Grille" in some results. Same name, address and hours on your site, Google, Yelp and OpenTable.
6. Write a Spanish page if you serve Spanish speakers. Spanish answers used Spanish media and travel blogs, and a Spanish menu page with the city and state is a source they can use.
7. Measure the same questions monthly on ChatGPT, Gemini and Claude, with dated screenshots, and keep Google and Yelp reviews healthy: diners check them before they book.

What we'd do for you, and what we don't promise

We do this work for US restaurants, remotely from Dubai, with traditional SEO included and ChatGPT and Gemini (Google AI) tracked on every plan. It starts with a free AI check: 10 real guest questions and the places AI names instead of you. The AI Visibility Audit (30 questions, 4 AI tools, 3 competitors, screenshots and a 90-day plan) is USD 549. Monthly plans are USD 1,249 (Foundation), USD 2,349 (Growth) and USD 4,399 (Authority), month to month, with no lock-in and no setup fee. Details at adjoltz.com/us.

In the first 30 days of any plan, if AI does not name you for at least 3 commercial questions on the list you approved, or you are unhappy for any other reason, we refund 100% of what you paid, in cash, within 7 working days (guarantee rules). We don't promise a top spot on any assistant, rankings, coverage in any publication, paid places on "best of" lists, covers or bookings, or a US office. We do not have one.