AI & Restaurants
How AI Assistants Find Restaurants
AI assistants learn about restaurants by four routes: what their search crawlers have indexed, what they fetch live when a user asks, what was in their training data, and — increasingly — what they can ask a connected service through a tool. For a restaurant, the routes that matter most are the live ones, because menus, hours and availability change and an assistant that relies on an old copy gives guests old answers.
Route 1: search crawlers
Assistants that answer with web results use their own search index or a search engine's. OpenAI says its OAI-SearchBot is used to show websites in ChatGPT's search features; Anthropic describes Claude-SearchBot as indexing content to improve the relevance of search results. If these crawlers cannot read your pages, your restaurant is less likely to be cited when someone asks.
Route 2: live fetches for a user
When a user asks about a specific place, an assistant may open the page there and then. OpenAI's ChatGPT-User and Anthropic's Claude-User do this on a user's request rather than crawling automatically; OpenAI notes that, because the user started the action, robots.txt rules may not apply. A page that loads quickly and states its facts as text is easier to read in that moment.
Route 3: training data
Separate crawlers collect content that may be used to train models — GPTBot at OpenAI and ClaudeBot at Anthropic. Training data is a snapshot: useful for general knowledge, unreliable for today's menu or tonight's opening hours. Both companies let site owners allow or block training crawlers independently of their search crawlers, in robots.txt.
Route 4: listings, maps and tools
- Business listings and maps carry your hours, address and photos; on Google, local results depend on relevance, distance and prominence.
- Directories and booking or ordering services can be connected to an assistant as tools. The Model Context Protocol is an open standard for such connections: the assistant asks the service for current data instead of guessing.
What you can decide
| Choice | Where | Effect |
|---|---|---|
| Allow search crawlers | robots.txt | Pages can be indexed and cited in answers |
| Allow or block training crawlers | robots.txt, per crawler | Content may or may not be used for model training |
| Keep listings consistent | Every listing you control | Assistants see one set of facts |
| Offer a live connection | Your menu or booking provider | Assistants can check current menus and availability |
Turning these choices into a checklist: how to make your restaurant ready for AI search.
How Thmenu restaurants are found
The robots.txt of menu.thmenu.com lets crawlers read restaurant pages, keeps them out of the technical API paths, and states a preference that the pages be used for search and AI answers but not for model training — the menus belong to the restaurants. If a restaurant chooses to be listed, which is off until it switches it on, assistants connected to Thmenu's MCP server can search the listed restaurants, read the live menu and check open reservation slots.
See how Thmenu works with AI assistants.
Sources
- OpenAI — Overview of OpenAI crawlers (accessed 28 October 2026)
- Anthropic Help Center — Does Anthropic crawl data from the web, and how can site owners block the crawler? (accessed 28 October 2026)
- Google Business Profile Help — Tips to improve your local ranking on Google (accessed 28 October 2026)
- Model Context Protocol — What is the Model Context Protocol (MCP)? (accessed 28 October 2026)