How an assistant builds its answer
When a guest asks for “a bistro in Chicago open on Sunday night” or “a Sunday roast near King’s Cross”, the assistant does not settle for what it learned in training. Most of the time it searches the web, much as you would, then writes a summary with links.
The providers document this themselves (pages consulted in November 2026):
- Google explains that AI Overviews and AI Mode run several searches related to the question, then show links to the pages they used. To appear, a page simply has to be indexed and eligible to show in Google Search with a snippet.
- OpenAI says its OAI-SearchBot crawler is what surfaces websites in ChatGPT search, and that sites which block it are not shown in ChatGPT search answers.
- Anthropic says Claude-SearchBot crawls the web for Claude’s search results, and that blocking it may reduce a site’s visibility in those results.
- Perplexity describes PerplexityBot as the crawler that surfaces and links websites in its results.
So the assistant reads what is public and indexed: business profiles, websites, review sites, press, guides. If those sources disagree, it gets things wrong, or picks someone else.
What you control
1. The same details everywhere
Name, address, phone number, opening hours, type of cuisine: write them exactly the same way on your website, your Google profile, your social accounts and every directory you appear in. One different opening time from page to page is enough to cast doubt. Remember holiday closures too: our guide to bank holidays and one-off closures covers them.
2. A complete Google Business Profile
Your business profile is a public page you fully control, so make sure it is right. A precise category, current hours, photos, a menu, a booking link: our article on getting bookings from your Google Business Profile goes through every field.
3. Your menu as text, not just a photo
A menu posted as a photo or a scanned PDF is hard to read. Publish it as text too, on a page of your website: dishes, prices, vegetarian or vegan options, allergens if you list them. A guest asking for “a restaurant with good vegan options” will only find you if those words are written somewhere. Guests look for the same thing before they book: see what to put online about your menu and prices.
4. One clear page with the essentials
Your website needs a page that answers at a glance: where, when, what, and how to book. The address as text, the opening hours, the cuisine, the booking link. Our guide to adding a booking widget to your website shows how to add booking to it, and what guests look for on a restaurant website tells you what to put first.
This is where gonnabook.it gives you a direct booking link to put on your website and your Google profile: those are the pages an assistant can read, and the guest who follows the link books with you, with no middleman.
5. Schema.org structured data
Structured data describes your restaurant in a format machines understand without guessing. The schema.org vocabulary has a Restaurant type with, among others, the address, opening hours (openingHoursSpecification), cuisine (servesCuisine), menu (hasMenu) and reservations (acceptsReservations). Google’s help on local businesses requires the name and address, and recommends the menu, cuisine, opening hours, phone number and website address.
To be honest about it: Google states that no special structured data is required for its AI features. This markup guarantees nothing; it prevents misreadings, provided it says exactly what the page shows.
6. A robots.txt that lets the right crawlers in
Your website’s robots.txt file tells crawlers what they may read. Blocking “AI” wholesale can also take you out of their search results. Tell search crawlers apart from those used to train models.
| Crawler | Provider | What it is for |
|---|---|---|
| Googlebot | Google Search, including AI Overviews | |
| Google-Extended | A token with no crawler of its own: Gemini training and grounding; no effect on Google Search | |
| OAI-SearchBot | OpenAI | ChatGPT search |
| ChatGPT-User | OpenAI | Visiting a page at a user’s request |
| GPTBot | OpenAI | Model training |
| Claude-SearchBot | Anthropic | Claude search |
| Claude-User | Anthropic | Visiting a page at a user’s request |
| ClaudeBot | Anthropic | Model training |
| PerplexityBot | Perplexity | Perplexity search results |
| Perplexity-User | Perplexity | Visiting a page at a user’s request |
| Applebot | Apple | Search in Siri, Spotlight and Safari |
| Applebot-Extended | Apple | A token with no crawler of its own: Apple model training |
If you want to be cited, allow at least the search crawlers. Blocking training crawlers is your call; according to the providers’ documentation, it does not remove your site from their search results.
What you do not control
Let’s be clear: nobody can guarantee you a recommendation. Answers vary from one question to the next, from one day to the next and from one assistant to another. Providers change their models and sources without notice. Reviews, press articles and guides weigh on the answer, and you do not write them.
So be wary of offers promising to put you “at the top of ChatGPT”.
How to test yourself
Once a month, take ten minutes:
- Ask two or three assistants the questions your guests ask: “[cuisine] restaurant in [your town]”, “where to eat on Sunday night in [your neighbourhood]”, “restaurant with a terrace or patio near [a landmark]”.
- Note whether you appear, and above all what is said about you: hours, address, prices, cuisine.
- When something is wrong, find where it comes from (often an old page or a forgotten directory listing) and fix it there.
- Ask directly “How do I book a table at [your restaurant]?” and check that the link offered is the right one.
Frequently asked questions about AI and restaurants
Do I need a website to be recommended by AI?
Not necessarily: an assistant can rely on your Google profile, reviews or articles. But without a website, you control no page that brings together your menu as text, your hours and your booking link. One clear page is often enough to correct mistakes that assistants pick up elsewhere, and to give them the right link for booking a table.
Do I need an llms.txt file to be cited by AI?
Not for Google: its help states that no new machine-readable files, AI text files or special markup are needed to appear in AI Overviews and AI Mode. The llms.txt file is a proposal published on llmstxt.org, not a standard. For a restaurant, a clear page with the address, hours, menu and booking link is far more useful than any extra file.
Does blocking GPTBot remove my restaurant from ChatGPT?
No, according to OpenAI: GPTBot is used for model training, and blocking it only signals that your content should not be used for that. ChatGPT search relies on a different crawler, OAI-SearchBot. If you want to appear in its search answers, allow OAI-SearchBot; you can still block GPTBot separately if you prefer to opt out of training.
Sources
- Google Search Central, AI features and your website, consulted in November 2026.
- Google Search Central, Google’s common crawlers, Google-Extended, consulted in November 2026.
- Google Search Central, Local business (LocalBusiness) structured data, consulted in November 2026.
- OpenAI, Overview of OpenAI Crawlers, consulted in November 2026.
- Anthropic, Does Anthropic crawl data from the web, and how can site owners block the crawler?, consulted in November 2026.
- Perplexity, Perplexity Crawlers, consulted in November 2026.
- Apple, About Applebot, consulted in November 2026.
- Schema.org, Restaurant, structured data vocabulary.
- llmstxt.org, The /llms.txt file, format proposal, consulted in November 2026.