llms.txt: a curated map for agents — not a proven citation switch

An /llms.txt file is a short Markdown overview of your site, meant to help software agents find the right pages without scraping navigation and ads. It is a community proposal, not a W3C or IETF standard.

Checked against the proposal site and Google’s Search Central docs on 13 October 2026. Other vendors change pages; follow the source links before you rely on a claim.

What the proposal actually specifies

According to llmstxt.org (v2), a conforming file:

Lives at /llms.txt (or under a path such as /docs/llms.txt, covering that path)
Starts with a required # heading (site or project name)
May include a blockquote summary, then free prose
Uses ## sections whose lists are links: - [Name](url): optional note
May point to clean Markdown twins of pages (for example page.html.md)

The proposal’s own framing: agents are expected to read the file on demand, then follow links. It is described as complementary to robots.txt and sitemaps, not a replacement.

Who has said they use third-party llms.txt?

Party	What we can verify	Source
Google Search (incl. AI Overviews / AI Mode)	Does not use llms.txt (or similar “AI text files”) for visibility or rankings. Google says maintaining one for other systems is fine and neither helps nor harms Search.	Google AI optimization guide
OpenAI crawlers	Official bot docs tell site owners to control GPTBot, OAI-SearchBot and ChatGPT-User with robots.txt. That page does not document reading third-party llms.txt. OpenAI does publish an llms.txt-style index for its own developer docs.	OpenAI bots; proposal notes labs’ own docs files
Anthropic / Perplexity	Publish llms.txt for their own documentation (per the proposal). Public crawler guidance we checked centres on robots.txt; we found no vendor statement that their consumer assistants use other sites’ llms.txt as a citation input.	llmstxt.org; Anthropic/Perplexity crawler pages linked from our AI crawler guide
Documentation platforms	Several docs hosts generate llms.txt automatically (listed on the proposal site). That is about publishing docs for coding agents, not about ChatGPT search rankings.	llmstxt.org Integrations

Honest takeaway: a good llms.txt is a tidy, agent-friendly summary. It is not a documented lever that makes ChatGPT, Claude or Perplexity cite you. Do not buy or sell it as a citation guarantee.

A sample that matches the proposal shape

Adapt names and URLs. Our free check only looks for structure (a # heading plus ## sections), not whether any assistant will read the file.

# Example Bakery Ltd

> Neighbourhood bakery in Copenhagen. Sourdough, cakes and catering. Not a chain.

Human-readable site; this file is a short map for software agents.

## Pages

- [Home](https://www.example.com/): Opening hours and today’s specials
- [About](https://www.example.com/about): Legal name, address, contact
- [FAQ](https://www.example.com/faq): Allergens, catering, wholesale

## Optional

- [LinkedIn](https://www.linkedin.com/company/example): Company page

How this fits our Month-1 pack

We still draft an llms.txt because it is cheap, matches a published format, and gives humans and tools one place to see your entity facts.
We never claim it changes AI citations. Our free check labels the absence as a structural gap, not as “you will not be cited”.
Prioritise crawlable pages, accurate About facts, and robots rules for AI search bots over polishing llms.txt. See the AI crawler guide and structured data guide.

Check your own file

Run the free AI-readiness check