How to create a useful llms.txt file
What is an llms.txt file?
llms.txt is an experimental proposal for a concise Markdown file at a site's root that helps language-model tools find high-value pages and descriptions. A practical file includes one H1 title, a short blockquote summary, H2 sections, and absolute Markdown links with brief notes.
It is not a robots control, sitemap replacement, or established ranking signal. Publishing the file does not guarantee that a crawler will fetch it, that a model will use it, or that a product will receive citations or search visibility.
How to generate a useful llms.txt
Choose canonical public pages that explain the organization, product, documentation, pricing, research, and contact path. Favor a small curated set over every URL on the site. Describe what each linked page contains rather than filling the file with keywords.
Review every generated line, save the output as plain text at /llms.txt, return a successful text response, and keep links current. Sensitive, private, gated, or noncanonical URLs do not belong in the file.
- Use exactly one top-level title.
- Write a factual summary that identifies the site and audience.
- Group related canonical URLs under descriptive H2 headings.
- Use absolute HTTP or HTTPS links with concise notes.
- Update the file when the site's core information architecture changes.
What this llms.txt validator checks
The validator checks a useful baseline: one H1, a blockquote summary, at least one H2, absolute Markdown links, no raw HTML, and a file under 100 KB. These checks make the document easier to interpret and catch common publishing mistakes.
The validator does not fetch URLs, test server headers, verify factual accuracy, or predict support by any AI product. After publishing, open the live /llms.txt URL directly and confirm that the response is public, readable, and current.