In one sentence
llms.txt is a file for AI crawlers (GPTBot, Claude-Web, etc.) that conveys "this is how our site is organized" in Markdown. It is placed at the site root (/llms.txt).
What does this look like in practice?
For example, you create a file at https://example.com/llms.txt and write something like:
# Example Corp
> Example Corp provides B2B SaaS sales-support tools.
## Services
- [Product overview](https://example.com/product)
- [Pricing](https://example.com/pricing)
## Learning resources
- [User guide](https://example.com/docs)
With this, AI can quickly understand "these are the key pages of this site."
Why it matters (validated by GEO Meter data)
In GEO Meter's smoke analysis (the "travel for enthusiasts" topic, 20 domains), 40% of the top AI-citation cluster had llms.txt deployed, versus 10% for the bottom cluster (+30pp gap). The same +20pp gap was observed for the "fat reduction" topic, with the trend confirmed consistently across multiple topics.
- Minimal implementation cost (done in 1-2 hours)
- Quick to take effect (reflected in AI citation behavior within weeks)
- Few competitors yet (per GEO Meter observations, adoption among Japanese companies is still limited)
This is one of the measures with the highest ROI in GEO.
How to write it
Minimal structure:
# Site name
> 1-2 line site description
## Key pages
- [Page name](URL): concise description
For details, see the llms.txt full implementation guide (to be published).
Common mistakes
- Writing HTML (Markdown is correct)
- Too long (rough target: 2,000-5,000 characters, emphasizing summarization)
- Updates lapse (weekly to monthly updates serve as a freshness signal)