llms.txt: an honest guide, with a working example
By the Motionexa Research Desk · last verified 2026-06-10
llms.txt is a Markdown file at your site root giving language models a curated map: one-paragraph definition, key facts, annotated links. Proposed by Jeremy Howard (Answer.AI, Sept 2024). Honest status in 2026: AI crawlers fetch it and adoption is growing, but no major engine guarantees it affects citations — treat it as an hour of cheap insurance plus a positioning exercise, paired with actual crawler access. Worked example and five writing rules included; this site's own file is at /llms.txt.
What llms.txt is
llms.txt is a plain-Markdown file served at your site root (/llms.txt) that gives language models a curated map of your site: what you are in one paragraph, then annotated links to the pages that matter, in an order you choose. The proposal comes from Jeremy Howard (Answer.AI) in September 2024 [1]. A companion convention, llms-full.txt, inlines the full text of key pages for models that want depth without crawling.
The honest status report (mid-2026)
Be clear-eyed about this: no major engine has committed to llms.txt as a ranking input. Adoption is a growing convention — strongest among developer-tool and AI companies (Anthropic, Mintlify-hosted docs, many devtools publish one) — and AI crawlers do fetch the file when present. Treat it as cheap insurance and a curation exercise, not a hack: it costs an hour, can't hurt, gives retrieval systems a clean self-description to lift, and the discipline of writing it usually exposes how muddled your positioning is. Anyone selling llms.txt as a guaranteed visibility lever is overselling a text file.
The format
Markdown, with a loose spec [1]: an H1 with your name, a blockquote summary, optional context paragraphs, then H2-grouped link lists with one-line annotations. An Optional section marks pages models can skip under length pressure.
A worked example for a B2B SaaS
# Acme Deploy > Acme Deploy is a deployment automation platform for > small engineering teams. Flat pricing, SOC 2 Type II, > founded 2024. Key facts: 5-minute setup; integrates GitHub/GitLab; plans from $49/month. ## Product - [How it works](https://acme.dev/how/): 90-second overview - [Pricing](https://acme.dev/pricing/): all plans, no "call us" - [Security](https://acme.dev/security/): SOC 2, data handling ## Comparisons - [Acme vs Octopus Deploy](https://acme.dev/vs/octopus/) - [Acme vs Jenkins](https://acme.dev/vs/jenkins/) ## Optional - [Changelog](https://acme.dev/changelog/) - [Blog](https://acme.dev/blog/)
Five rules that make it useful
- First paragraph = your category definition, phrased the way a buyer would ask. This is the sentence you want quoted.
- Curate ruthlessly — 10–25 links beats 200. It's a reading list, not a sitemap (you already have a sitemap).
- State facts machines can lift: pricing, compliance, founding year, integrations — concrete and dated.
- Keep it in sync — a stale llms.txt confidently feeding old pricing to a model is worse than none. Ours regenerates with every deploy.
- Pair it with crawler access — the file does nothing if robots.txt blocks GPTBot, ClaudeBot and PerplexityBot (checklist items 06–12 in the audit checklist).
You can read this site's own implementation at /llms.txt — we eat the cooking.
Questions people ask
Q.01 Do ChatGPT, Perplexity or Google actually read llms.txt?
AI crawlers fetch the file when present, and some retrieval pipelines use it as a site summary — but as of mid-2026 no major engine has publicly committed to it as a ranking or citation input. Honest framing: a low-cost, growing convention that gives models a clean self-description, not a proven lever. The hour it takes is justified by insurance value and the positioning clarity it forces.
Q.02 What's the difference between llms.txt and robots.txt?
robots.txt is access control — which crawlers may fetch which paths. llms.txt is editorial guidance — what your site is and which pages best explain it. They're complementary: robots.txt opens the door, llms.txt hands the visitor a reading list.
Q.03 What's llms-full.txt?
A companion file that inlines the complete text of your key pages, so a model can ingest your essential content in one fetch without crawling. Useful for documentation-heavy products; for a typical marketing site, a tight llms.txt plus crawlable pages is enough.
Q.04 Where should the file live and in what format?
Plain Markdown served as text at https://yourdomain.com/llms.txt — site root, no authentication, no redirect chain. Subdirectories or CMS attachments defeat the convention since tools look only at the root path.
Sources & further reading
- [1] Jeremy Howard / Answer.AI, the llms.txt proposal and specification, September 2024 (llmstxt.org)
- [2] Anthropic and major devtools documentation sites publishing llms.txt implementations, 2025–2026
- [3] OpenAI / Anthropic / Perplexity crawler documentation on fetch behavior
Want this analysis run on your category? The full audit — 40+ prompts, 5 engines, scorecard, source map, fix worksheet — is a flat $1,200, with the founding-client evidence guarantee.