The llms.txt format in 30 seconds
# Acme Running
> Acme Running sells trail and road running shoes online and ships to the EU and US.
Prices are in EUR. All shoes have a 60-day return window.
## Products
- [Trail shoes](https://acme.example/trail/): Full trail range with sizes and prices
## Optional
- [Blog](https://acme.example/blog/): Training guides- H1 with the site or project name — the only required element.
- Blockquote with a one-sentence summary. Write it the way you’d want an AI to describe you.
- Free text for key facts: markets, prices, policies, what you don’t do.
- H2 sections of Markdown links, each with an optional
: description.
What llms.txt does — and doesn’t — do
Be clear-eyed about it. llms.txt was proposed by Jeremy Howard of Answer.AI in September 2024 to help models use websites at inference time. Developer tools and some AI agents read it, and many documentation sites publish one. But Google has said it does not use llms.txt in Search, and none of the major AI search products document it as a signal for citations.
What actually decides whether ChatGPT, Perplexity or AI Overviews can cite you is the basics: crawlers allowed, content in the HTML, and pages that answer questions clearly. Check those first with the AI crawler checker and the full audit.
How to write a good llms.txt
- Keep it short: the 10–30 pages that answer the questions customers actually ask.
- Describe each page by the question it answers, not by marketing copy.
- Use absolute URLs that return 200 and are not blocked in robots.txt.
- Put archives and secondary pages under
## Optional. - Update it when you add or retire important pages.
Publishing it
Upload the file so it is served at https://yoursite.com/llms.txt with a text/plain or text/markdown content type and a 200 status. The audit checks both.