Skip to content
Akkrev
Get started

What gets a hotel, restaurant or café recommended by AI? What the research says

Updated By Akkrev

What independent studies show about how ChatGPT, Gemini and Google AI choose hotels, restaurants and cafés, what the evidence doesn't support and what to do.

Key points

  • AI recommends the places it can find, read and see confirmed by other sources. There is no paid route into the answer.
  • Most sources AI cites for local businesses are websites and listings the business can manage. Yext counted 86% across 6.8 million citations.
  • Being talked about matters more than backlinks. Ahrefs found brand mentions correlate far more strongly with AI visibility than links do.
  • Review volume moves the needle. ChatGPT named 84% of hotels with 2,500+ Google reviews and 52% of those with under 50.
  • Each AI reads different sources. Google's AI Mode cited Google's own pages 99.7% of the time for hotels. ChatGPT leaned on TripAdvisor and Booking.com.
What the research points to
  1. 1

    Be readable

    AI crawlers allowed, menus and rooms as text, facts on your own site.

  2. 2

    Be consistent

    Same name, address, hours and details on Google, Bing, Apple and booking sites.

  3. 3

    Be described for the question

    Say the things guests ask about: dog-friendly, parking, vegan, private dining.

  4. 4

    Be talked about

    Recent reviews, local press, guides and mentions on other sites.

  5. 5

    Measure, then repeat

    Ask the same guest questions every week and watch which sources AI cites.

AI assistants recommend the hotels, restaurants and cafés they can find, read and see confirmed by other sources. The research so far points to five things that matter most: an AI can read your information, your facts agree everywhere, you are described for the exact question a guest asks, other people talk about you, and you have plenty of recent reviews. Nothing in the evidence suggests you can pay your way into an AI answer.

There is a lot of guesswork sold as AI search advice. This page sticks to what independent studies and the AI companies themselves have published, and says where the evidence is thin.

Why this matters now

AI is now a normal way to choose where to eat and stay. BrightLocal’s 2026 survey of 1,002 US adults found 45% had used an AI tool to get a local business recommendation, up from 6% a year earlier. 31% had used ChatGPT and 23% Google’s AI Mode. AI is now the third most used source of local recommendations, behind Google and Facebook.

People still check. 97% of those AI users said they at least sometimes check the answer against real reviews. So AI decides who makes the shortlist, and reviews decide who gets the visit.

1. Most of what AI cites, you can manage

The largest public study of AI citations comes from Yext. It looked at 6.8 million citations across ChatGPT, Gemini and Perplexity in mid 2025. 44% pointed to business websites and 42% to listings such as Google, Yelp and directories. Reviews and social made up 8%. Forums like Reddit were 2%.

A larger follow-up covering 155.5 million citations in early 2026 found 80% still pointed to websites or listings.

Yext sells listings management, so read this with that in mind. But the pattern is consistent with other studies: for questions tied to a place, AI leans on sources built to describe places.

What it means for you: your website, Google Business Profile, Bing Places, Apple Business Connect, Yelp, TripAdvisor and booking listings are the raw material. If they are incomplete or disagree, AI has less to work with.

2. Each AI reads different sources

Local Falcon ran 66,240 AI searches for 998 US hotels in 2026 and recorded every source cited.

AI Named the hotel Most cited sources
Gemini 64.6% Booking.com, Expedia, HotelPlanner
Google AI Mode 64.5% Google’s own pages (99.7%)
ChatGPT 63.5% TripAdvisor (27.1%), Booking.com (23.1%)
Google Maps 56.4% Google listings
Google AI Overviews 10.4% No Overview shown in about half of searches

Hotel websites themselves made up only about 7% of the sources ChatGPT and Gemini cited.

For restaurants, Local Falcon found Google’s own listings made up 48.6% of the sources Google’s AI Overviews cited, and the top 10% of US restaurants took 74.5% of all AI visibility.

What it means for you: being strong on Google helps with Google’s AI. Being strong on TripAdvisor, Yelp and booking platforms helps with ChatGPT. You need both, and you need to check each assistant separately.

Ahrefs studied 75,000 brands across ChatGPT, AI Mode and AI Overviews. The strongest signals were mentions: YouTube mentions correlated at about 0.74 and branded web mentions at 0.66 to 0.71. Backlinks and the number of pages on a site correlated far more weakly.

These are correlations. They show that brands people talk about get named more often, not that one press mention will get you named. But they explain why local guides, food writers, travel press and “best of” lists keep showing up in hospitality answers.

What it means for you: a mention in a local guide, a newspaper review or a respected blog does more for AI visibility than another page on your own site.

4. Review volume matters more than star rating

In Local Falcon’s hotel study, ChatGPT named 52.1% of hotels with fewer than 50 Google reviews and 83.8% of hotels with 2,500 or more.

A 2026 study of 4,776 restaurants, cafés and bars in Bali found the same pattern. Venues with more reviews and more outside mentions were much more likely to be named. Star rating made no difference to whether a venue got named at all.

Reviews also carry detail. A guest who writes “quiet room, great breakfast, easy parking” gives AI three reasons to match you to three different questions.

5. Specific, sourced content gets cited

The original academic study on this topic, GEO: Generative Engine Optimization, was presented at KDD 2024 by researchers from Princeton and others. Across 10,000 test queries, adding statistics, quotations from credible sources and clear citations raised a page’s visibility in AI answers by up to 40%. Stuffing pages with keywords did not help.

What it means for you: “great food and friendly service” gives AI nothing to quote. “Two AA Rosettes, 40 covers, tasting menu £85, vegetarian menu on request” does.

6. No special tricks are required

Both Google and OpenAI say the same thing in their own documentation.

  • Google: there are no additional requirements and no special structured data needed to appear in AI Overviews or AI Mode. A page must be indexed and eligible to show a snippet in normal Search.
  • OpenAI: any public website can appear in ChatGPT search. Do not block its OAI-SearchBot crawler. It ranks results using multiple factors and placement is not guaranteed.

So llms.txt files and special markup are not shortcuts. Being indexed, readable and well described is the foundation.

What the evidence does not support

  • Paying for placement inside the answer. OpenAI says ads do not influence ChatGPT’s answers. See can you pay ChatGPT to recommend you.
  • Guaranteed rankings. Answers change daily. Anyone promising a fixed position in ChatGPT is guessing.
  • Fake reviews. They break platform rules, are illegal in the UK under the DMCC Act, and are removed from the sources AI reads.

Where to start

  1. Ask ChatGPT, Gemini and Google 5 to 10 questions your guests really ask, without your name in them.
  2. Note who gets named and open the sources each answer cites.
  3. Make sure your facts are complete and identical on your website and every listing those answers use.
  4. Make sure your website and listings say the specific things guests ask about.
  5. Ask for reviews steadily, and look for a local guide or writer who should know about you.
  6. Ask the same questions again in a few weeks and compare.

Akkrev samples your guest questions every week across models from the ChatGPT, Gemini, Claude and Grok families and shows who gets named and which sources are cited. See plans, or read why AI recommends your competitors.

Questions

What actually makes ChatGPT recommend a business?

OpenAI says ChatGPT ranks search results using multiple factors to find relevant, reliable information, and that placement is not guaranteed. Independent studies show the businesses it names are usually well covered on listings and review sites, have plenty of reviews and are described for the exact question asked.

Do I need special schema or an llms.txt file to appear in Google's AI?

No. Google says there are no extra requirements and no special structured data needed to appear in AI Overviews or AI Mode. A page must be indexed and eligible to show a snippet in normal Google Search.

Is it better to get backlinks or mentions?

For AI visibility, mentions. In Ahrefs' study of 75,000 brands, branded web mentions correlated with AI visibility at roughly 0.66 to 0.71, well above backlinks. It is a correlation, not proof that mentions cause visibility.

Does star rating decide who AI recommends?

Less than you would think. A 2026 study of 4,776 venues in Bali found review count and outside mentions predicted being named, while star rating made no difference to getting named at all.

Can a small independent venue beat a chain?

Yes, on specific questions. Chains win broad questions because they have more coverage. Independents win when they are clearly described for a dish, an area, an occasion or a need that the chain does not mention.

Sources

Related