People ask ChatGPT, Perplexity and Google's AI Overview where to eat. Those answers are built on completely different things from what you may have been optimising your website for over the past few years.
I've done redesigns of more than twenty Danish restaurant websites over the past year. Here's what the AI models look for, and what's usually missing.
What restaurant AI search is
When someone types "find a good Italian restaurant in Aarhus with an outdoor terrace" into ChatGPT, the model gathers information from several sources at once, reviews, websites, menus, and produces one concrete answer. Not ten links.
Traditional SEO is about ranking high so people click through. AI search is about being cited and recommended directly in the answer, often without the user ever visiting your website.
Why old-school SEO alone isn't enough
AI models look for:
- Structured, easily understood information about hours, menu, prices and location
- Consistent details across Google Business, Tripadvisor and the website
- Recent reviews with concrete detail
- Natural language that answers actual questions
- Authority, meaning whether other credible sources mention you
At one franchise chain, the meta description was "default description" on all 14 pages and there was no JSON-LD anywhere. At a Chinese restaurant in Aalborg, the whole Wix site was on noindex without the owner knowing. At one Aalborg restaurant, the site's entire indexable text came to 226 words. The AI had almost nothing to read.
Your website is only one data point
Which means you can't do one thing and be finished. At a French bistro, the external sources contradicted each other about opening hours, and the restaurant's own pages didn't list them at all. At a restaurant in northern Denmark, the hours page and the booking page disagreed. The model can't resolve that conflict for you.
What to prioritise
- Structured data: Restaurant schema for menu, hours and reviews
- Consistent details everywhere, name, address and hours identical down to the last character
- Natural language in the content: write "crispy duck breast with port wine sauce", not just "duck breast"
- A steady flow of reviews with concrete detail, which means asking guests to be specific
- FAQPage schema with eight questions answered word for word from your own material
- Authority through mentions in local media and on relevant lists
What it does to your analytics
Some restaurants see organic clicks from Google fall while reservations rise. People get the answer straight from the AI and call or book without visiting the website. That sounds alarming if you only look at visitor counts, but it's a sign it's working, just measured differently.
What I don't promise
No "number one in AI answers", no guaranteed citation. What I do is make your data consistent and machine-readable, and remove what's making you invisible.
Frequently asked questions
What's the difference between AI search and ordinary Google search?
Google gives a list of links you click through yourself. AI search gives a concrete answer with recommendations, often without you visiting the underlying website.
Should I drop traditional SEO?
No. Good traditional SEO, a fast site, relevant content, decent links, is still the foundation. AI search builds on top of it with structured data and consistency.
How long before I see results?
The first signs typically show within two to three months. Full effect takes four to six months, depending on how much needs fixing.
Does this only work for big chains?
No. Smaller local restaurants often do better, because AI models reward specific, genuine information over anonymous chain language.
What role does social media play?
A bigger one than most people think. Updated, active profiles signal that you're still relevant right now. At one French bistro, Instagram wasn't linked from the website at all.
One last thing
I don't do brand awareness. I do revenue, rankings and websites that sell. Check your opening hours across platforms today, fix the discrepancies, and ask for three specific reviews this week.