The essentials
- Definition: llms.txt is a Markdown text file placed at a site's root (/llms.txt) that describes the site and lists its most useful pages, for the benefit of large language models.
- Origin: proposed in September 2024 as an open standard; adopted by thousands of sites and by documentation publishers, without any formal commitment from OpenAI, Google or Anthropic.
- Contents: a title, a summary, sections of annotated links to your reference pages; an llms-full.txt version can carry the full text.
- Its use in 2026: cheap, risk-free, useful for the tools that read it; it replaces neither indexing, nor page structure, nor crawler permissions.
The llms.txt file is a text file, written in Markdown and placed at a site's root (https://www.exemple.fr/llms.txt), that introduces the site to large language models: what it is, what it offers, and which pages to read first. Proposed in September 2024 by Jeremy Howard as an open standard, it answers a limitation of LLMs : their context window is small, and HTML pages, loaded with menus, scripts and adverts, are hard to read. llms.txt offers a clean, condensed way in.
What an llms.txt contains
| Element | Format | The full destination URL |
|---|---|---|
| Title | A Markdown H1, required | # GreenRed |
| Summary | A short blockquote | > Plateforme de suivi marketing digital pour PME… |
| Description | Free paragraphs | Your business, your audience, what the site covers |
| Link sections | An H2 then a list of links with a comment | ## Resources then - [KPI marketing](url) : les indicateurs à suivre |
| Optional section | An "Optional" H2 for secondary links | Pages the model can skip when context is short |
A variant, llms-full.txt, carries the full text of your key pages in Markdown, for tools that prefer to load all the content at once. Technical documentation sites (frameworks, APIs) use it widely.
llms.txt, robots.txt and sitemap: three files, three roles
| A file | Audience | Role | Status |
|---|---|---|---|
| robots.txt | Every crawler | Allowing or forbidding crawling | A recognised standard (RFC 9309) |
| sitemap.xml | Search engines | Listing URLs and their modification date | A recognised standard |
| llms.txt | AI assistants and agents | Introducing the site and pointing to the useful pages, in Markdown | An open proposal, adopted voluntarily |
llms.txt forbids nothing and allows nothing: AI crawler access is set in robots.txt. Nor does it replace the sitemap: it selects and annotates, where the sitemap lists exhaustively.
What AI does with it in 2026
None of the major publishers (OpenAI, Google, Anthropic, Perplexity) has announced that it systematically reads llms.txt for its answer engine, and Google has said it doesn't use it for search. The tools that do use it are mainly agents and development assistants, which load it when a user gives them a site's address, plus some specialist crawlers. Its direct effect on citations in ChatGPT or the AI Overviews is not demonstrated. Its value lies elsewhere: it costs an hour, carries no risk, forces you to state clearly what the site offers and to designate your reference pages, and it is ready should the engines adopt it. The mechanisms that really decide whether you get cited remain indexing, crawler permissions and page structure, described in indexing by AI and structuring your pages for AI.
Creating your llms.txt
- Write the title and the summary: the company's name, then one sentence saying what it does, for whom, and where.
- Describe the site in a few paragraphs: your offers, your audiences, the themes you cover, the useful facts (founding date, location, specialisms).
- Select your reference pages: offer pages, About, content hubs, pillar guides, definitions; 20 to 200 links depending on the size of the site, each with a one-line description.
- Generate the file: by hand for a small site, with a script or a CMS plugin otherwise; the full procedure is in creating your llms.txt.
- Publish it: at the root, as plain text (MIME type text/plain or text/markdown), reachable without any blocking.
- Verify: open the URL, paste it into an AI assistant and ask it to summarise the site; the summary you get should match what you want said about you.
Mistakes to avoid
- Treating it as a magic lever: an llms.txt won't get a site cited if its pages are badly structured or not indexed.
- Putting everything in it: 2,000 links with no comments amounts to a sitemap; select and describe.
- Letting it age: dead links and discontinued offers give a false picture to the tools that read it.
- Blocking AI crawlers alongside it: a welcoming llms.txt with a robots.txt that forbids GPTBot is a contradiction.
How GreenRed helps
Rather than juggling several tools, GreenRed's GEO and AI module brings these metrics together in a single dashboard, compares them over time and tells you which actions come first. You can try it free, with no card, from the Pricing.
Frequently asked questions
What is the llms.txt file?
A Markdown text file, placed at a site's root, that introduces the site to large language models and lists its most useful pages with a description. Proposed in 2024 as an open standard, it aims to give AI assistants a condensed, readable way in.
Does llms.txt improve visibility in ChatGPT or Google?
There is no direct evidence in 2026: Google says it doesn't use it for search, and the major assistant publishers haven't announced reading it systematically. It is mostly read by agents and development tools. Visibility depends on indexing, AI crawler permissions and page structure.
What is the difference between llms.txt and robots.txt?
robots.txt controls crawler access (allowing or forbidding crawling) and is a recognised standard. llms.txt controls nothing: it describes the site and points to the useful pages, for AI, on a voluntary basis.
Should a small business create an llms.txt?
Yes, because it costs an hour, carries no risk, and forces you to state your offer and your reference pages clearly. Keep it up to date and expect no direct effect on citations; the main effort goes into page structure and entity consistency.