Skip to the content.

Real-world survey

Generated on 2026-10-21 by scripts/real_world.py against 19 public llms.txt files. Sites change their files, so rerun for current results.

llms.txt Errors / warnings Findings Broken links
https://llmstxt.org/llms.txt 0 / 0 clean 0 of 3
https://docs.anthropic.com/llms.txt 0 / 156 link-notes-format×149, empty-section×4, prose-in-section×2, missing-summary×1 not checked of 808
https://docs.stripe.com/llms.txt 0 / 59 prose-in-section×45, duplicate-link×7, list-item-not-link×5, missing-summary×1, empty-section×1 not checked of 468
https://vercel.com/llms.txt 0 / 2 empty-section×2 0 of 28
https://svelte.dev/llms.txt 0 / 1 empty-section×1 0 of 7
https://docs.cursor.com/llms.txt 0 / 6 empty-section×5, missing-summary×1 0 of 0
https://developers.cloudflare.com/llms.txt 0 / 1 missing-summary×1 not checked of 112
https://supabase.com/llms.txt 0 / 1 missing-summary×1 0 of 41
https://docs.perplexity.ai/llms.txt 0 / 0 clean not checked of 170
https://hono.dev/llms.txt 0 / 1 link-notes-format×1 not checked of 94
https://bun.sh/llms.txt 0 / 2 missing-summary×1, prose-in-section×1 not checked of 321
https://mintlify.com/llms.txt 0 / 5 prose-in-section×5 0 of 41
https://fastht.ml/docs/llms.txt 0 / 0 clean 0 of 21
https://modelcontextprotocol.io/llms.txt 0 / 1 missing-summary×1 not checked of 362
https://docs.expo.dev/llms.txt 0 / 7 duplicate-section×4, empty-section×3 not checked of 565
https://orm.drizzle.team/llms.txt 0 / 17 empty-section×9, list-item-not-link×8 not checked of 446
https://docs.github.com/llms.txt 0 / 2 prose-in-section×2 not checked of 117
https://nextjs.org/docs/llms.txt 0 / 6 prose-in-section×5, empty-section×1 not checked of 299
https://langchain-ai.github.io/langgraph/llms.txt 0 / 0 clean 0 of 43

Sites affected per rule

What the survey changed in llmstxt-lint

The first run of v0.1.0 over these files produced many errors that were really rule problems, not site problems. v0.2.0 fixes them:

v0.1.0 behaviour Real-world example v0.2.0
Every bullet in a notes-only section was an error (list-item-not-link) ## Notes in svelte.dev, ## When to use Vercel in vercel.com Note-only sections are reported once as empty-section; list-item-not-link is a warning and only fires in sections that contain links
H3 headings were warnings docs.expo.dev (54), docs.stripe.com (16), docs.anthropic.com (12) use H3 sub-groups H3 allowed; only H4+ warns
No detection of HTML docs.cursor.com/llms.txt 308-redirects to cursor.com/docs, which returns HTML unless Accept includes text/markdown (plain curl gets HTML) New html-content error; the built-in fetcher sends Accept: text/plain, text/markdown, */*
--check-links (new) flagged HTTP 401 supabase.com lists its MCP endpoint https://mcp.supabase.com/mcp, which requires auth 401/403/429 are treated as reachable, not broken
llms-full.txt checked like llms.txt hono.dev/llms-full.txt starts with <SYSTEM>…</SYSTEM> and contains many H1s Dedicated full mode: only requires an H1 somewhere

Notable remaining findings

Compatibility with the reference parser

llms_txt.parse_llms_file (PyPI llms-txt 0.0.7, from AnswerDotAI/llms-txt) raised an exception on 5 of the 19 files: svelte.dev, vercel.com, orm.drizzle.team, docs.anthropic.com and docs.stripe.com. The cause is a non-link bullet inside an H2 section (upstream issue #16). It also fails on a file with a title but no > summary line directly followed by an H2. llmstxt-lint reports these cases as empty-section / list-item-not-link / missing-summary, so authors can stay compatible with strict consumers.