llms-full.txt - the entire documentation bundled into one file for AI
Tests whether you publish /llms-full.txt, a concatenated Markdown dump of your primary documentation.
What this signal tests
We check for /llms-full.txt - the larger companion to /llms.txt. Where llms.txt is a curated table of contents, llms-full.txt is the full text of your primary documentation concatenated into a single Markdown file. The convention originated with Mintlify and has spread to several other docs platforms. It lets an AI fetch your entire knowledge base in one HTTP request.
Why it matters for your visibility in AI
Modern long-context models can hold hundreds of thousands of tokens at once. When an AI is asked a deep question about your product, it can either crawl your site page by page (slow, expensive, often blocked) or fetch a single llms-full.txt and load the whole knowledge base into context. The second path is dramatically faster and gives much more accurate answers - because the model sees everything at once rather than guessing which page to read next. The consequence for a documentation-heavy business: better technical answers in AI assistants, fewer hallucinated APIs, and a much higher chance that when a developer asks ChatGPT "how do I do X with your product" they get a correct, current answer instead of a 2023 guess. For SaaS and developer tools especially, this is one of the highest-leverage AI-readiness changes you can make.
Pass criteria at a glance
| Criterion | Passes when |
|---|---|
| 200 + size >1024 bytes. |
How we test it
We send GET /llms-full.txt and check that the response is 200 and the body is larger than 1KB (a sanity check that the file isn't an empty placeholder). We do not currently validate the Markdown beyond size, because real-world implementations vary in structure - some include the H1 of every source page, others a single H1 and flat content. Both pass. Content type should be text/markdown or text/plain.
Show technical detection method
GET /llms-full.txt; 200 + content >1KB.
If your site fails: how to fix it
- Generate the file from your existing documentation. If you use Mintlify, this is on by default. Docusaurus, Nextra, and VitePress have community plugins.
- If you have a hand-built docs site, write a build-time script that walks your /docs Markdown source, strips frontmatter, and concatenates the files with separator headings.
- Cap the file at a size that fits comfortably in modern context windows - roughly 1–5 MB of Markdown is the sweet spot. Larger than that and most models will only ingest the start.
- Serve it at exactly /llms-full.txt with Content-Type: text/markdown; charset=utf-8.
- Regenerate on every docs deploy so it stays in sync. See the conventions at https://llmstxt.org/.
Quick facts
| Maturity | EMERGING |
|---|---|
| Weight | medium |
| Category | Emerging Standards |
Primary sources
Related signals
Frequently asked questions
Is this part of the official llms.txt spec?
It is an extension convention popularised by Mintlify alongside llms.txt, not a separate formal spec. The llmstxt.org site documents it as a sibling pattern. There is no central authority; the convention has spread by adoption.
How big should it be?
Big enough to be useful (more than a kilobyte), small enough to ingest. For most sites a few hundred kilobytes to a few megabytes of Markdown is right. If your full documentation is huge, consider publishing per-section files (llms-product.txt, llms-api.txt) in addition.
Should I include marketing pages?
Usually no. The convention is to include documentation, reference material, and substantive guides - the stuff someone asking a technical question would want. Marketing copy adds noise and burns context.
Is there an easier alternative?
Not really. Even with a great llms.txt, a model still has to fetch each linked page separately. The single-file shortcut is the whole point. Generated automatically by your docs build, the cost is one-time.
Run your own scan
Run a free scan and see how your site grades across all 155 AI-readiness signals.