llms.txt and llms-full.txt: what they are and whether you need them
llms.txt is a Markdown file at the site root: a name, a short description and a list of key pages with links. Jeremy Howard proposed it in September 2024. The file is cheap, but do not expect much from it.
Why it matters
- A short list of what matters is easier for an agent than crawling menus and footers.
- Real requests are rare. According to Ahrefs, 97% of sites with a valid llms.txt got zero requests to it in a month, and Google Search does not use it. That is why llms.txt weighs 3 out of 100 in our score.
- For API documentation the file is more useful: coding agents read it noticeably more often.
What to do
- 1
Follow the format
An H1 with the name, a quote with a one- or two-sentence description, H2 sections with links and notes. Link to .md versions of pages if you have them.
# Product name > What it is and who it is for, in one or two sentences. ## Product - [Features](https://example.com/features.md): what it does - [Pricing](https://example.com/prices.md): prices and limits ## Developers - [API](https://dev.example.com/): REST API, token auth - 2
llms-full.txt depends on the site
Documentation benefits from the full text in one file. For a marketing site, redirecting /llms-full.txt to the Markdown version of the homepage is enough, and we count that as a pass.
- 3
Serve a file, not HTML
A common mistake: an SPA returns its HTML shell with a 200 on /llms.txt. For an agent that is not a file.
curl -sI https://example.com/llms.txt | grep -i content-type
What the scanner checks
llms.txt exists
A short file for language models: what the product is and where the key pages are.
llms-full.txt exists
The full text of the site or docs in one file. An agent reads it in a single request.
llms.txt is structured well enough to navigate
A heading, sections and at least five links. A one-line file does nothing for an agent.
Check your own site
Paste the address and get the same breakdown in half a minute.
Sources
Other guides
- An agent can open your site: no CAPTCHA, no JavaScript wall, an honest 404
- robots.txt for AI: YandexAdditional, GPTBot, ClaudeBot and Content-Signal
- Markdown for AI agents: .md page versions and Accept: text/markdown
- sitemap, metadata and JSON-LD: how an agent learns who you are
- API and MCP for agents: OpenAPI, MCP Server Card, OAuth discovery