llms.txt in one minute
llms.txt is a plain-text file at the site root (https://your-domain/llms.txt), written in Markdown. It is a site guide for models: who you are, what you offer, which pages matter, and how the company should be described.
The problem is practical. When AI fetches a site it meets nav, pop-ups, and slogan copy, and struggles to extract a clean core. llms.txt is the summary you hand over.
Usual format
Community convention looks like this:
# Company or site name
> One sentence on what you do, with the core terms.
A short extra paragraph—scope, regions, credentials.
## Key pages
- [Page title](full URL): what this page answers
- [Page title](full URL): what this page answers
## FAQ
- [Question](full URL): short answer
How a company site should write it
1. Put facts in the lead sentence, not slogans
“Industry-leading one-stop digital solutions” tells a model nothing. “Founded in 2023 in Shenzhen; GEO and AI-search semantic monitoring; 500 companies served” has density.
2. List only the pages that matter
Do not paste the sitemap. Pick 10–20 pages that represent the business: products, core solutions, flagship cases, authoritative notes. One sentence after each link: what question that page answers.
3. Put frequent questions in the file
What users ask AI is often what sales hears every day. Those Q&As in llms.txt raise the odds of being cited.
Caveats
llms.txt is not yet a required standard on every AI platform. It is a forming convention. Cost is near zero, risk is none, and several products already say they read it. For a company, that is a high-leverage move.
It does not replace on-page quality. It is a map, not the content. If the pages it points to are vague, the effect stays limited.