llms.txt is a proposed plain Markdown file, placed at the root of a website, that gives large language models a short, curated map of the site’s most useful pages. A good workflow drafts it from your sitemap and analytics, edits it by hand, publishes it at /llms.txt, and updates it on a schedule. It is cheap to produce, but it is a convention rather than a standard that search engines have committed to.
What is llms.txt, exactly?
The idea was proposed in 2024 by Jeremy Howard of Answer.AI. The reasoning is simple: websites are full of navigation, scripts and ads, and models have limited context windows. A short Markdown file can point a model straight to the pages that explain what a company or project does.
The proposed format is light. It starts with an H1 containing the site or project name, then a short blockquote summary, optional paragraphs of context, and H2 sections with lists of links, each with a one-line description. An “Optional” section can list secondary links that a model may skip. Some sites also publish llms-full.txt, a longer file with full page content in Markdown.
Who supports it?
Adoption has been strongest among developer documentation sites and software tools, and several documentation platforms can generate the file automatically. At the time of writing, major search engines have not said they use llms.txt for ranking or for choosing sources in AI answers, and Google representatives have publicly downplayed it. Treat it as low-cost preparation, not as a ranking lever.