G-LabG-Lab Studio

Research · Answer · 17 September 2026

What is llms.txt, and does a restaurant need one?

llms.txt is a plain-text Markdown file at the root of a website that tells AI models what the site is and where the important pages are. It can help an assistant describe a restaurant. It does not let an assistant book one: in our Amsterdam sample, 34 restaurant websites served an llms.txt in September 2026, up from 9 in July, and none of them offered a way to book.

9

restaurant websites served an llms.txt, July 2026

34

served one in September 2026

0

of those offered a way for an assistant to book

What the file is

llms.txt is a proposal from 2024, not an official web standard. The file lives at /llms.txt, is written in Markdown, and usually holds a one-paragraph summary of the site followed by links to its key pages. It is meant for language models the way robots.txt is meant for crawlers, except that it describes instead of restricting.

Whether a particular assistant reads it is that assistant's decision. Our data cannot tell you which ones do, and we do not claim it.

What we saw on restaurant websites

When we checked the working websites in our sample in July 2026, 9 served an llms.txt. When we re-checked them in September, 34 did, roughly one in five. In the same check, the number offering a booking interface an assistant could call was still zero.

That is a good instinct aimed at the wrong layer. The file improves how accurately a restaurant can be described. It does nothing for whether an assistant can use it.

Is llms.txt worth it for a restaurant?

It is cheap, so there is little reason not to have one, as long as it is accurate and kept current: opening hours, cuisine, address, how to book. Just do not treat it as the step that makes you bookable by AI. That takes something an assistant can act on.

What an llms.txt for a restaurant can contain

A short, factual file is enough:

  • # Restaurant name, then one paragraph: cuisine, neighbourhood, what it is known for.
  • Opening hours and address, as plain text.
  • Links to the menu page and the booking page.
  • How bookings work: party sizes, how far ahead, deposits if any.
  • Contact for groups and events.

What actually makes a restaurant bookable by an assistant

Availability and booking offered as data: a documented API, an MCP server or another interface with a schema, so an assistant gets a typed answer to three questions. What do you offer, when are you free, can you take this booking now. In our sample, no restaurant website offered this.

Questions

What is llms.txt?

A plain-text Markdown file at /llms.txt that summarises a website for AI models and links to its key pages. It is a 2024 proposal, not an official standard.

Is llms.txt actually used?

Some sites publish it and some AI tools may read it, but reading it is up to each assistant, and our research does not measure that. What we measured is that the number of Amsterdam restaurant websites in our sample serving one went from 9 in July to 34 in September 2026, and none of them became bookable by an assistant.

Does llms.txt make my restaurant bookable by ChatGPT or other assistants?

No. It describes the restaurant. Booking needs an interface an assistant can act on, such as an API or an MCP server, or at least a booking form an agent can operate.

Is llms.txt the same as robots.txt?

No. robots.txt tells crawlers what they may fetch. llms.txt describes the site and points to its important pages.

Sources, all published with method and aggregate data: Being found isn't being booked (September re-run) · Amsterdam Restaurant Agent-Ready Booking Index 2026 · How do you test whether an AI agent can book a restaurant?.

Disclosure: G-Lab also builds G-Guest, a booking system with an interface an assistant can call. The study names no venues and is measured the same way for everyone.

To cite: G-Lab Research (2026). What is llms.txt, and does a restaurant need one? https://g-lab.studio/research/answers/llms-txt-for-restaurants

More answers

G-Lab Research studies how local businesses are found, understood and booked by people and by AI assistants. We publish the method before the findings, name every source, and version the methodology so a result can always be traced to the way it was measured.

Questions about a figure, the sample or the method: hello@g-lab.studio. Corrections are published in the methodology changelog. Quoting us: how to cite.