How to get your restaurant recommended by ChatGPT

Illustration — a restaurant picked and recommended by ChatGPT among many others

A couple of tourists walk out of Le Bon Marché and pull out a phone: "A good Mediterranean restaurant near here, for tonight?" Two years ago, that question went to Google Maps. Today it increasingly goes to ChatGPT — which doesn't answer with ten links, but with two or three addresses, each with a short summary. The question that should keep every restaurateur up at night: why those three, and not you?

How ChatGPT picks a restaurant

There's no magic and no lottery. When asked for a local recommendation, an AI like ChatGPT (or Claude, Gemini, Perplexity) searches the web and cross-checks several signals:

  • What already ranks on Google. The AirOps study (March 2026) measured it: Google's first position is cited 3.5 times more than the following ones in ChatGPT's answers. Your local SEO remains the foundation.
  • What its crawlers can read. A slow website, built from images or blocked to AI crawlers, is invisible. A fast, well-structured website in real text is a source of choice.
  • What is precise and verifiable. Cuisine, neighbourhood, price range, hours, terrace: the AI recommends what it can state without risk of being wrong.
  • What others say. Google reviews, platforms, local press: the AI cross-references. A consistent reputation carries real weight.

In other words: the AI doesn't pick the "best" restaurant. It picks the one whose information is the most reliable and the easiest to cite. Which is excellent news for serious independents.

The 6 levers, from most urgent to finest

1. A website AI crawlers can read

The absolute prerequisite. A real website (not just Instagram), fast, in clean HTML, open to OpenAI's, Anthropic's and Google's crawlers. If your current site is a scanned PDF inside a 2017 Wix page, everything else is pointless — this is the core of my restaurant website work.

2. Complete restaurant structured data

Schema.org "Restaurant" markup tells machines, in their own language: cuisine, address, hours, prices, reviews, booking link. Invisible to your customers, decisive for AI.

3. An llms.txt file

This small file at the root of your website introduces your establishment directly to AI models: who you are, your specialities, your practical information. Very few restaurants have one — a head start worth taking now.

4. Factual content, citable as-is

"Mediterranean brasserie in the 7th, mains €18–28, terrace, continuous service on weekends." That's a sentence an AI can repeat without getting it wrong. Your story, your specialities, your answers to frequent questions (groups, gluten-free, private hire) — every factual page is a chance to be cited.

5. Regular reviews and a consistent reputation

AI reads your Google reviews and cross-checks with platforms. Volume matters less than regularity and your replies. (Bonus: AI can actually help you reply to every review without giving up your evenings.)

6. A Google profile perfectly aligned with the website

Same hours, same address, same phone number, everywhere. The slightest inconsistency lowers the algorithm's trust — and the AI moves on to the next place. The Google profile + website duo remains the bedrock.

Run the test tonight

Open ChatGPT and ask it the questions your customers ask: "good [your cuisine] restaurant [your neighbourhood]", "where to eat near [metro station]", "restaurant with a terrace [district]". Look at who gets cited, and what the AI says about you — or doesn't. That's exactly the starting point of AI search optimization (GEO): a diagnostic of what AI answers today, then a plan to change the answer.

The window is ideal: almost no restaurant is working on its AI visibility in 2026, and Google's AI Overviews arrive in France this summer. The first places to settle in will be hard to dislodge.

Frequently asked questions

Can ChatGPT really recommend my restaurant?

Yes. When asked for a place to eat, ChatGPT draws on the web: pages that rank well on Google, fast websites its crawlers can read, structured information and reviews. An independent restaurant with a factual, well-built website can be recommended ahead of chains — brand size matters less than information reliability.

Do I have to pay OpenAI or Google to be listed in AI answers?

No. There is currently no paid programme to appear in the organic recommendations of ChatGPT, Claude, Gemini or Perplexity. Everything rests on the quality of your online presence: website, structured data, reviews, consistency of information. That is exactly what AI search optimization (GEO) is about.

How long before I see results?

The technical foundations (structured data, llms.txt, speed) are picked up as soon as the crawlers come back — days to weeks. Progress in the recommendations themselves is then gradual: it follows your Google rankings, review momentum and the consistency of your presence. You measure by regularly asking the AIs the questions that matter to you.

Want to know what ChatGPT already says about your restaurant? Request a diagnostic — I'll tell you honestly, examples included, and we'll decide the next step together.