What is llms.txt?

llms.txt is an optional file placed in a website’s root directory that summarises, in Markdown, what the site is about and where its most important pages are for language models.

The proposal was published by Jeremy Howard on llmstxt.org in September 2024. The idea is that a model can learn what is where from a single file without crawling the whole site. It is a proposal, not a document from a standards body; supporting it is voluntary.

Its place in search is clear: Google Search does not use it, and Google states in its own documentation that no special file is needed for its AI features. Its value is as a table of contents for agents and tools that read the site directly.

Format

The file is plain Markdown, and the proposal makes only the first-level heading mandatory. The rest is optional but comes in a set order:

• First-level heading — the name of the site or project; the only required element.
• Blockquote — a one- or two-sentence summary of the site.
• Free paragraphs — notes on how to read the file.
• Link lists under second-level headings — each line carries a link and an optional short note.
• A section headed “Optional” — links that can be skipped when a shorter context is needed.

The proposal also describes serving a clean Markdown version of each page at the same address with .md appended, so a model can read the text without the navigation and decoration. llms-full.txt, which gathers the linked detail into a single file, is not part of the proposal but a convention that has spread. This site keeps both side by side: llms.txt as a short index and llms-full.txt as the full directory listing every guide, term and tool.

Example

1. # Example Ltd.
2. > Istanbul-based manufacturer of technical fabrics; products, data sheets and certificates.
3. ## Products
4. - [Knitted fabrics](https://www.example.com/products/knitted/): weight and width table
5. ## Optional
6. - [About us](https://www.example.com/about/)

Common mistakes

• Treating the file as a lever for Google rankings — search does not use it; the real work is in the pages themselves.
• Writing information into the file that is not on the site — the file should summarise the pages; diverging information erodes trust.
• Writing relative links — if every line carries a full address, the file can be read on its own.

You can create your own file with the llms.txt generator. How the file relates to robots.txt and what goes into it is covered in the llms.txt and robots.txt guide; its measured effect is in the 2026 data article.

If you want to know how your brand is cited in ChatGPT, Google AI Overviews and Perplexity, our GEO consulting records which questions and engines cite you and builds the prerequisites.