GEO for restaurants & cafés: Get recommended by ChatGPT

The question "where should we eat tonight" used to end in a list of ten places and a decision made by the guest. Increasingly it ends in a chat window, and the decision arrives pre-made: two or three restaurant names, each with a line of reasoning — cuisine, distance, whether the kitchen is still open. The restaurant that is named gets the table. The restaurant that is not named is not being rated badly; it simply never enters the conversation, and there is no second page of results where it could still turn up.

How a guest actually asks

The prompts are shorter and more specific than the search queries they replaced. "Which is the best Italian restaurant in Cologne?" is one of them. "Where can I eat Thai food in Cologne with proper vegan options?" is the more revealing of the two, because it carries a filter the assistant has to check against something. Guests also ask for a place near a landmark they have just named, for a kitchen that is still serving late, or for somewhere that works for a group of twelve. Every one of those questions is answered from a fact — opening hours, cuisine, a dish list, a reservation policy — and an assistant can only use a fact it finds in a form it can read. A homepage sentence saying the kitchen is open "until late" is not such a form.

Why restaurants are hit especially hard

Hardly any other industry has as many intermediaries sitting between the business and the guest. Reservation platforms hold the booking, delivery portals hold the menu, review sites hold the reputation, city guides hold the description. Each of them publishes a version of your restaurant that is structured, consistently formatted and easy for a machine to read — which is exactly why assistants quote them. Your own website is often the least machine-readable description of your own house, and when the portal entry is out of date, so is the answer the guest receives. The second effect is the shortlist itself: nobody compares fifteen restaurants at half past seven on a Friday. Two or three names are the entire decision, and the guest never learns which places were left out.

What AI Rank Suite measures

The visibility check does not start with content. It asks the assistants 25 questions in their own voice — phrased the way a guest phrases them, for your city and your cuisine — and scores every answer for one thing: is your restaurant named? What comes back is a score plus the evidence behind it: which places were named instead, in what context, and which of your pages the answers were built from. Everything runs on your own API key, so the check can be repeated after every change and you can see whether the change moved anything. This is a measurement of the answers as they stand today, not a position you can buy.

The ready-made restaurant pack

You do not have to invent any of this yourself. The restaurant pack ships with the industry's real material and applies it in one click: it fills the entity's subject areas — cuisine, menu, reservations, catering — sets up the schema types that restaurant answers are built from, and installs the prompt set together with the FAQ block. The schema types in the pack are:

• Restaurant — the business itself, with address, opening hours and cuisine
• Menu — the dishes as data instead of a PDF
• FAQPage — the questions guests ask before they book
• Review — the ratings, on your own site and not only on a portal

The FAQ block starts from the four questions that actually decide a booking: "Do you take reservations?", "Do you have vegetarian and vegan dishes?", "Is there parking nearby?" and "Can you cater for allergies?" All of it is a starting point, not a verdict. You edit what does not fit, delete what you do not offer, and add what is specific to your house — the Sunday closing day, the terrace, the set lunch. The pack saves you the blank page; the accuracy stays your decision.

Measure first, then decide

The free edition sets up in about a minute and needs no AI key at all: schema, the FAQ block, llms.txt and crawler control. That alone makes your restaurant readable to the assistants that already answer these questions. Pro adds the AI features for $99 a year — the 25-question visibility check and the writer for the answer pages you are missing. Agency is $199 a year for unlimited sites, which is the version that makes sense if you look after several houses. We do not promise you a place in an answer, and it is worth distrusting anyone who does. What you get is a measurement, the structured facts underneath it, and the ability to run the same check again next month and compare.