Free tool · Generator
Hand the assistants a map of your site.
llms.txt is a plain-text index at the root of your domain that tells a model what is here and what matters. This builds one from your sitemap.
01Why it matters
A crawler guesses your structure. A map removes the guess.
An assistant arriving at a site infers its shape from navigation, internal links and a sitemap written for search engines. That inference is usually adequate and occasionally badly wrong, particularly on sites where the important pages are not the most linked ones.
llms.txt is a proposed convention for saying it outright: a markdown file at the root listing the pages worth reading, with a line describing the site. It is not a standard anybody is obliged to honour, and it is cheap enough that the asymmetry favours publishing one.
This builds a starting point from your sitemap, titled from your URLs. Edit it before you publish, because the value is in the curation rather than the completeness, and a file listing every page in sitemap order is just the sitemap again.
02Questions
- What is llms.txt?
- A plain-text, markdown-formatted file published at the root of a domain that lists the pages an AI assistant should read and describes what the site is. It is a proposed convention rather than a ratified standard, intended to do for models roughly what robots.txt does for crawlers.
- Do assistants actually read it?
- Support is partial and evolving. Some tools look for it, others ignore it entirely, and no engine guarantees it. It costs one file to publish, which is why the trade-off usually favours having one even while adoption is uneven.
- Where do I put the file?
- At the root of your domain, served as plain text: yoursite.com/llms.txt. Same placement convention as robots.txt.
- Should I list every page?
- No. The value is curation. A file that mirrors your entire sitemap tells a model nothing your sitemap did not already say, so trim it to the pages you would want quoted.
Next
A map helps once the pages are readable.
An llms.txt pointing at pages a model cannot parse, or is not allowed to fetch, does not help anybody. The full check tells you whether the pages behind the map hold up.