TLDR
- We asked eight AI models where to eat and drink, again and again. TripAdvisor and Google were level in the sources they cited; the other two-thirds is a mix of guides, blogs, booking sites and venues' own websites.
- Your Google Business Profile and local "best-of" lists are two gatekeepers you can work on directly.
- No venue is an AI favourite yet: most venues the models named were cited by only one of them, so the field is still open.
AI restaurant recommendations come from many sources, and we asked eight AI models where to eat and drink 506 times in Q2 2026 to see which. Most probes ran in Portuguese cities, with smaller comparison samples from Manchester and three Spanish cities. TripAdvisor and your Google listing came out level. The other two-thirds is a mix: local guides, food blogs, best-of lists, booking sites and venues' own websites.
Your Google Business Profile is one source you manage directly. Local "best-of" lists are another you can earn your way onto.
There is no established AI favourite yet. Most of the venues AI named were cited by only one of the eight models. The field is open for restaurant and bar owners, as long as AI can find you.
Who Does AI Actually Listen To When Recommending Your Place?
AI listens to a mix of sources: TripAdvisor and Google came out level, and the other two-thirds came from guides, blogs, booking sites and venues' own websites. When a diner asks ChatGPT or Gemini, "where should I eat and drink tonight?", that AI assistant pulls from sources it trusts. Between May and June 2026, we ran 506 probes across eight AI models and 11 cities. This is a Portugal-led study: most cities are in Portugal, plus Manchester and three Spanish cities. We then grouped every cited source into four groups:
- The Aggregator: TripAdvisor.
- Everything else: Bloggers, city food guides, "best independent restaurants in {city}" lists, booking sites such as TheFork and OpenTable, and venues' own websites.
- Google: The search results that come directly from your Google Business Profile.
- Social: Instagram and YouTube.
Where Do AI's Restaurant and Bar Recommendations Really Come From?
TripAdvisor and Google supplied the same share of what AI cited. Social supplied far less. The rest is spread across many kinds of site. The shares are set out in the last section of this post.
The most-cited single hosts in that mix were the Michelin Guide and Time Out Portugal. Booking sites were a visible part of the mix too. We did not measure how many came from venues' own websites. TripAdvisor and Google, the listing you control, ended up level.
So there are two things to work on. Give local writers a reason to talk about your place, and give Google a profile that is clear enough for AI to repeat.
One more finding about Google. When ChatGPT's search mode gave us links in our probes, 114 of the 142 were Google Maps listings. The rest were venues' own websites, and not one was a TripAdvisor page. The page it most often hands to the diner is your Google Business Profile. If that listing is incomplete, that is what a potential customer sees.
Do the results hold in the UK and Spain?
We cannot say yet, because most of our probes happened in Portuguese cities and the UK and Spanish samples are small. In our sample, TripAdvisor's share of citations was close in the Portuguese and Spanish probes and lower in Manchester. Read that as a first look only. The sample sizes and shares are in the last section.
One thing stood out:
- Of every TripAdvisor citation in our sample, 63% pointed to tripadvisor.com.br, the Brazilian edition. Most of those citations came from our Portuguese probes. With so few Manchester and Spanish probes, we draw no line between the countries.
How Can Restaurant and Bar Owners Influence AI Recommendations?
Two of these sources are within your direct influence: your Google Business Profile and the local writers who cover your area.
- Your Google Business Profile — the lever you fully control. This is your most direct path to AI visibility. Choose the most specific category that fits, such as "Neapolitan Pizzeria" for a pizza place. Fill in every field, as Google's own advice on local ranking says to give complete and detailed business info, and reply to your reviews. This is the one source on our list you own outright, and it feeds the share of citations that went to Google.
- Local best-of lists and bloggers. Guides and best-of lists, such as the Michelin Guide and Time Out Portugal, were the most-cited hosts in the mix. We did not measure the exact share of guides and blogs. Invite local food writers to your venue. Pitch your story to city guides. Give them an angle to feature your restaurant or bar.
- TripAdvisor still counts. Keep your listing accurate and answer reviews promptly. It supplied as many citations as Google, one source among several.
Is There Still Time to Get Your Restaurant or Bar Noticed by AI?
AI has not picked its winners yet. Of the venues named across our 506 probes, most were named by only one of the eight models, and only a small number reached "consensus," meaning they were named by four or more models. The counts are in the last section.
What Do the Venues Most AI Models Named Have in Common?
Two of the three were each named by six of the eight models, and both have a Google listing rated 4.7 or 4.8 stars (July 2026). We looked at the top of that consensus list, at the venues' live Google Business Profiles and the sources AI cited. Three of them, for restaurant and bar owners to read:
- Zé Manel dos Ossos — a traditional bistro in Coimbra, in Portugal. It was named by six of the eight models. When we checked its Google listing in July 2026 it showed a budget average spend and 4.7 stars from a very large number of reviews.
- Cervejaria Ramiro — named for cities it is not in. This seafood institution in Lisbon, in Portugal, was named for Lisbon, and also for Porto and Setúbal, both in Portugal.
- A Cozinha por António Loureiro — a fine-dining room in Guimarães, in Portugal. It was named by six of the eight models. It is in the north of Portugal, with 4.8 stars on its Google listing in July 2026.
How can you see how your Google listing compares with the venues near you?
Run a Competitor Check on your venue. A Competitor Check is a one-off audit of your Google profile and reviews, set beside your top local competitors. And that is where booteek helps: it shows where the venues near you are ahead of your listing, so you know which part to work on first.
What's the Bottom Line for Restaurant and Bar Owners on AI Search?
Keep your Google Business Profile complete and current, and give local writers a reason to name you. In BrightLocal's 2026 survey of US adults, 45% said they used AI tools for local recommendations. In our Q2 sample, guides, best-of lists and Google listings were among what AI cited.
booteek is the AI agent for independent restaurant and bar owners. We work through your Google Business Profile with you, so AI assistants name your venue.
Keep your profile fresh, and be the name they find.
Where do these figures come from?
Every figure here comes from our Q2 2026 probes (public AI visibility snapshots, 5 May to 25 June 2026) unless it names another source.
| Who AI cited | Share of the 844 publisher citations |
|---|---|
| TripAdvisor | 14% |
| Google (fed by your Google Business Profile) | 14% |
| Social (Instagram, YouTube) | 6% |
| Everything else: local guides, food blogs, best-of lists, booking sites and venues' own websites | the other two-thirds |
The Michelin Guide and Time Out Portugal were the most-cited single hosts, 25 citations each. Booking sites added 81 citations.
| Sample (Q2 2026) | TripAdvisor share of citations |
|---|---|
| Portugal (7 cities, 159 sourced probes) | 15% |
| UK — Manchester (18 sourced probes) | 6% |
| Spain — Santander, Bilbao, Vigo (12 sourced probes) | 17% |
- Of the venues named across the 506 probes, more than four in five were named by only one of the eight models. Only 94 venues were named by four or more.
- Two of the three venues above had Google listings in July 2026 with 4.7 stars across more than 3,000 reviews (Zé Manel dos Ossos) and 4.8 stars across 667 reviews (A Cozinha por António Loureiro).
How we counted: 506 probes across 8 models and 11 cities, from 5 May to 25 June 2026. Source shares are of the 844 publisher citations the search-enabled models returned; Gemini's internal grounding-redirect URLs were excluded as they aren't publishers. This was a bounded sample, not a census — so read the figures as "in our Q2 sample." And one honest limit: we measured what AI cites, not what actually fills tables. A citation isn't a booking, after all.
