Foundational
llms-full.txt: when to publish one and what goes inside
tl;dr
llms-full.txt is the long-form sibling of llms.txt. It concatenates your key content into one markdown file so an AI assistant can ingest your site in a single fetch. Worth doing for docs sites, knowledge bases, and any site under ~500 pages.
// your move
See what AI assistants say about your site.
Free, 30 seconds, no signup. We'll check all 15 signals and tell you which ones to fix first.
Run the auditWhat's the difference vs llms.txt?
llms.txt is the table of contents. llms-full.txt is the book.
llms.txt links out to each page. llms-full.txt pastes the actual content together with clear headings between each entry, so an assistant can answer questions without making 40 follow-up requests.
Should you publish one?
Yes, if you run a docs site, knowledge base, API reference, or a marketing site under about 500 meaningful pages.
Probably not, if you run a 50,000-page e-commerce catalog or a news site with thousands of new articles a week. The file gets too big to be useful and too painful to keep fresh.
How to assemble llms-full.txt
- 01Make a list of the URLs you want included. Use the same list as your llms.txt as a starting point.
- 02Fetch each page and convert it to clean markdown. Tools like Mozilla's Readability + Turndown, or services like Firecrawl, do this well.
- 03Strip nav, footers, cookie banners, ads, and any boilerplate. The file should be content only.
- 04Concatenate the pages. Put a level-2 heading with the original URL between each one so context survives. Example: ## https://yoursite.com/docs/auth
- 05Save as llms-full.txt and upload to your site root.
- 06Test the URL in a browser. Then test the file size — under 1MB is comfortable, under 5MB is workable, over that and most assistants will sample or skip.
- 07Set a calendar reminder to regenerate it monthly, or wire it into your deploy pipeline.
“Manual assembly works once.”
Automating it
Manual assembly works once. After that, automate. A small script that reads your llms.txt, fetches each link, converts to markdown, and writes the bundle is roughly 50 lines in any language.
If you're on a static site generator, you can build llms-full.txt as part of the build. Astro and Next.js both let you write a tiny route that produces the file on each deploy.
Common questions
How big can llms-full.txt be?
Aim for under 1MB. Most assistants will happily fetch up to a few MB, but the bigger the file the more it gets truncated when included in a context window.
What if my site is too big?
Publish llms-full.txt for your highest-leverage subset (docs, top guides, key product pages) and rely on llms.txt for the rest.
Does the order of pages matter?
Yes. Put the foundational explainers first, then specifics. Assistants summarize from the top down when the file is large.
Keep reading
// your move
See what AI assistants say about your site.
Free, 30 seconds, no signup. We'll check all 15 signals and tell you which ones to fix first.
Run the audit ⚡