AI visibility for restaurants
“Where should we eat tonight” is now a question people ask AI assistants, and it comes with follow-ups a host would recognize: gluten-free options, a table for eight, whether you take reservations, how loud it gets. Assistants answer from what they can read, which for restaurants means your site, your consistency across platforms, and, fatally often, whether your menu exists as text or as a photograph of text. The gap between being recommended and being invisible is usually one of three fixable problems.
What do diners ask AI?
Constraint-shaped questions, because that’s what choosing a restaurant is:
- “Best Italian near me that takes reservations tonight”
- “Restaurants in [city] with good gluten-free options”
- “Where can eight people eat near [venue] before a 7pm show?”
- “Is [restaurant] good for a date, and what does dinner cost?”
- “Does [restaurant] have vegan options?”
Each answerable constraint you publish as readable text is a query you can win. The restaurant whose site plainly says “we hold tables for parties up to ten, book by phone” gets relayed to the group organizer verbatim, and the one whose site says nothing gets skipped for somewhere the assistant could verify.
What makes restaurants invisible to AI?
Three culprits, all common. PDF menus and image menus lead the list: a scanned menu is a picture, AI crawlers read text, and a restaurant whose entire offering lives in a JPEG has published nothing as far as retrieval is concerned. Reservation-platform microsites come second, since many render through JavaScript or block crawlers outright, and for some restaurants that microsite is the only web presence.
Contradictions finish the job: when your site, your Google profile, and a directory disagree about hours, a model hedges, and hedging models recommend the place whose details agree everywhere.
What should your site actually say?
The facts a model needs to relay a recommendation: hours, address, phone, cuisine, price range, reservation policy, and the menu as HTML text with prices.
Add the constraint answers diners actually filter on, meaning dietary accommodations, group capacity, parking, and noise level, each stated in a plain sentence rather than implied by photos. Restaurant schema with hours and price range is worth the five minutes for Google’s rich results, with the usual honesty about mechanism: models read visible text, so the markup supplements the words rather than replacing them.
Which fixes matter most for restaurants?
Three, in order of how often they’re the whole problem. Get your content out from behind JavaScript, which for restaurants usually means replacing a PDF or widget menu with real HTML. Check your firewall and platform aren’t blocking AI crawlers, particularly if your main presence lives on a reservation platform’s subdomain.
And phrase your key headings as diner questions, because “Do you have gluten-free options?” with a first-sentence answer is exactly the passage an assistant lifts, while a heading that says “Cuisine” is not.
How does your restaurant score?
Run the free scan. It checks crawler access, whether your menu and details exist as readable text, response speed, and answer shape, then gives you an AI-written analysis of what matters most for your site. Every failed check links to a step-by-step fix you can hand to whoever manages your website, and the report link makes that handoff a one-liner. Selling products instead of tables? E-commerce has its own version of the readable-text problem.
See where your site stands. The free scan takes about fifteen seconds and shows every fix.
Run a free AI visibility scan