Real-world llms.txt examples

Verified directory

Every entry below was fetched and parsed by our checker on 2026-06-15: the file existed, returned plain text, and started with a valid H1. The check runs weekly; entries that disappear are removed automatically. Out of 215 hosts in our fixed measurement panel, 112 currently serve a llms.txt file.

Sections and links as measured in the live file. Raw data: adoption-stats.json · adoption-panel.json

Annotated examples

Six files worth reading line by line, with notes on what makes each one good. All links go to the source files on the original sites.

Looking for a larger running list? The community directories directory.llmstxt.cloud and llmstxt.site track sites that publish llms.txt. To check any specific site yourself, see how to find a website's llms.txt.

Patterns to copy

1. Group by product surface, not by content type. Stripe, Cloudflare and Anthropic all section by product (Payments, Workers, Claude API), not by “blog / docs / guides.” That matches how an LLM client usually queries.
2. One sentence per item, factual. Avoid marketing language. The note after the colon is for disambiguation, not for selling.
3. Keep the root file small. If you have many products, publish per-product variants (e.g. /workers/llms-full.txt) instead of one giant root file.
4. Pair with llms-full.txt. Most documentation adopters publish both. The map is for navigation, the full file is for ingestion.
5. Use absolute URLs. Every example we list does this. Relative URLs make the file ambiguous when copied around.

Submit an example

Want your llms.txt listed in the directory? The requirements are simple:

• The file is publicly accessible at the listed URL.
• It validates against our validator with no errors.
• You can describe it in two sentences without marketing language.

Reach out via the channel on the about page with your URL; if the file passes the checker, it joins the panel at the next weekly run.