People ask ChatGPT for a good restaurant now. If yours doesn't show up in that answer, you're losing guests you never knew existed. This isn't about tricking the model. It's about making your data so clear it can't misunderstand who you are.
I've done redesigns of more than twenty Danish restaurant websites, and most of them fail on the basics.
How ChatGPT picks a restaurant
Unlike Google, which shows ten links and lets you choose, ChatGPT has to make the choice for you. It points at one or two places. It pulls information from several sources at once: your website, your Google Business profile, reviews and mentions. The more sources that say the same thing about you, same name, same address, same kind of food, the more confident the model gets about recommending you.
Consistency is the whole game. Not tricks.
What's usually wrong
At one franchise chain, the front page was 96 words with no images, the meta description was "default description" on all 14 pages, and there was no JSON-LD anywhere. At a Chinese restaurant in Aalborg, the entire Wix site was set to noindex without the owner knowing. At a French bistro, the restaurant's Instagram wasn't linked anywhere, and sameAs in their JSON-LD was an empty list, so Google and AI had no verified connection between the domain and the profiles.
When the data is thin, inconsistent or invisible, the AI skips you.
What you need in place
- Google Business profile: filled in, verified, correct opening hours
- Consistent name, address and phone across the website, Facebook, Google and Tripadvisor
- Restaurant schema on the site with menu, prices and location
- FAQPage schema: the single best addition for AI search. Eight questions answered word for word from your own material, covering reservations, allergies, price range, group size and wheelchair access. Never invent answers here.
- Reviews with concrete detail, not just stars
- Mentions in third-party sources like local media and food blogs. AI trusts what others say about you more than what you say about yourself.
A robots.txt that explicitly allows OAI-SearchBot, GPTBot, ClaudeBot, PerplexityBot and Google-Extended, plus an llms.txt with facts and prices in markdown, helps too. On most sites I've measured, the AI crawlers already get HTTP 200. The door is open.
Avoid the aggregator trap
Directory sites pull data from a profile nobody maintains and fill in the rest with AI-generated, invented descriptions. If your own presence is weak, that slop is what ChatGPT leans on. Clean up your own sources so the model has something better to draw from.
What I don't promise
I don't promise ChatGPT starts citing you. Nobody can guarantee that. What I do is make your data consistent and machine-readable, so you have a real shot.
Frequently asked questions
How do I know whether ChatGPT already recommends my restaurant?
Ask it. Type "recommend a restaurant in your city serving your kind of food" and see whether you come up. Do it a few times, the answers vary.
Can I pay to be recommended by ChatGPT?
No. There's no paid advertising that directly influences which restaurants ChatGPT names. Only real credibility and good data work.
How long before changes take effect?
Typically four to twelve weeks, depending on how much needs fixing.
Do I need an English version of my website?
If tourists are part of your audience, yes. ChatGPT answers in the language the user asks in, and the more information it can find in that language, the better.
Does this work the same for small restaurants as for big chains?
Often better. Small restaurants with a clear niche and strong local reviews frequently outperform large chains with generic content.
One last thing
I don't do brand awareness. I do revenue, rankings and websites that sell. Clean up your core data this week, ask for three new reviews, and write one good FAQ straight from your own material.