H1 title
The name of the site or project. It is the only required element.
Guide · llms.txt
llms.txt is a Markdown text file at the root of a site that says in a few lines what the site is about and which pages to read to understand it. It is a 2024 proposal aimed at language models: it blocks and allows nothing, and Google has said it does not use it for Search.
llms.txt is an index of the site written for machines: a title, a summary and lists of links to the pages that matter, each with one line of explanation. It is cheap to make and does no harm, but it is not a lever for showing up on ChatGPT or Google: indexing, bot access and quotable content are. Where it really helps is technical documentation read by agents and coding assistants.
The proposal sets a simple Markdown format, in this order.
The name of the site or project. It is the only required element.
A block starting with the greater-than sign that explains in two or three sentences what the site is about.
Optional paragraphs with the context needed to read the rest correctly.
Each section groups links to the important pages, one per line, with a short description after a colon.
The section with this name holds the links that can be skipped when space is short.
A minimal file for a company site. GeoSuite's own is public at trygeosuite.it/llms.txt.
# Company name
> What the company does, for whom and where, in two sentences.
## Products
- [Main product](https://www.example.com/product): what it is for and who it is for
- [Pricing](https://www.example.com/pricing): current plans and costs
## Guides
- [Getting started](https://www.example.com/guide): the first steps
## Optional
- [About us](https://www.example.com/about)The four most common mix-ups.
robots.txt tells bots where they may go, llms.txt blocks and allows nothing. If a bot is blocked in robots.txt, llms.txt does not let it through.
A sitemap lists every page to index, llms.txt picks a few and explains what they are for.
Google has said publicly that it does not use llms.txt. No major AI engine has confirmed reading it to pick the sources of its answers.
It is a variant that puts the full text of the pages into a single file. It is mostly useful for technical documentation, which an agent can read in one go.
Half an hour of work, if the right pages already exist.
A few: the ones that explain who you are, what you sell, how it works and what it costs. An index with everything in it helps nobody.
After the link, one sentence that says what is on the page and which question it answers.
The file is served as plain text at /llms.txt on the main domain, reachable without a login.
When you add or remove an important page, update the file too: a broken link in an index is worse than no index.