Technical Signals · P1.4
AI-native formats: llms.txt, llms-full.txt, and markdown that bots actually read
tl;dr
Three AI-native signals: llms.txt (an index for assistants), llms-full.txt (the full content bundle), and markdown content negotiation (serve clean text when bots ask for it). Publishing all three takes an afternoon and is currently a competitive moat.
// 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 auditWhy these formats exist
The web was built for humans reading in browsers. AI assistants ingest pages by the thousand and waste tokens parsing nav menus and cookie banners.
These three formats are the community's answer: lightweight, predictable, machine-first. They sit alongside your regular site, not instead of it.
The three files at a glance
- ·/llms.txt — a markdown index of your most important pages. ~15 minutes to write.
- ·/llms-full.txt — your key content concatenated as one markdown file. ~2 hours, or automated.
- ·Markdown content negotiation — your server returns markdown when a bot sends Accept: text/markdown. ~30 minutes of config.
Which one to publish first
Start with llms.txt. It's the lowest-effort, highest-signal win. Even if no other site on your competitive list has one, you get noticed.
Add llms-full.txt next if you have a docs site, knowledge base, or under 500 important pages.
Layer content negotiation last. It's the most technical, but it future-proofs you as more assistants adopt the Accept header.
“Start with llms.txt.”
What 'good' looks like
Good llms.txt: every link works, descriptions are one sentence, structure mirrors how you'd give someone a tour.
Good llms-full.txt: stripped of chrome, under 1MB, regenerated on each deploy.
Good content negotiation: same URL serves HTML to browsers and markdown to bots. Vary: Accept set so caches behave.
Common questions
Is this an actual standard?
llms.txt is a community proposal that's getting picked up fast. Markdown content negotiation uses the same HTTP machinery you've used for decades. Neither is in an official W3C spec, but adoption is real.
Will publishing these hurt anything?
No. They're additive. Browsers ignore them. SEO doesn't care. You only gain.
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 ⚡