llms.txt generator in the llmstxt.org format

What llms.txt is, and is not

llms.txt is a proposed Markdown file at the root of a site that lists its most useful pages for language models. Think of it as a short, curated reading list that tells an AI model which pages explain your business best. It comes from the llmstxt.org proposal of September 2024. It is a proposal, not a standard: no major AI vendor has confirmed reading it, Google's AI guidance says it can be ignored, and it does not control access. robots.txt decides which bots get in; llms.txt, at most, offers a reading list.

Adoption is still small. In our robots.txt study of the Tranco top 10,000 (5,583 valid files), only 118 robots.txt files mention llms.txt at all (2.1%). Lighthouse now audits its format, so treat the file as a cheap hedge rather than a ranking factor.

The format in one example

# Example Company

> We make accounting software for small businesses in Singapore.

Short context paragraph that explains who the site is for.

## Product
- [Features](https://www.example.com/features): what the software does
- [Pricing](https://www.example.com/pricing): plans and prices

## Help
- [Getting started](https://www.example.com/help/start): first steps

## Optional
- [Company history](https://www.example.com/about/history)

1. One # heading with the site or project name. This is the only required line.
2. An optional blockquote with a one-line summary.
3. Optional paragraphs of context.
4. ## sections, each a list of links in the form - [Title](url): note.
5. An ## Optional section for pages a model can skip when space is short.

A finished file for a local business

# Riverside Physio

> Physiotherapy clinic in Riverside offering sports injury rehab and back pain treatment.

Open Monday to Saturday. Most treatments are covered by the main private insurers.

## Services
- [Sports injury rehab](https://www.example.com/services/sports): assessment and rehab plans
- [Back and neck pain](https://www.example.com/services/back-pain): treatment and exercise programmes

## Visit
- [Prices](https://www.example.com/prices): fees per session and package
- [Book and contact](https://www.example.com/contact): opening hours, address and phone

## Optional
- [Our team](https://www.example.com/team)

Every link has a note that says what the page answers, the summary says who the clinic serves and where, and the Optional section holds the one page a model can skip. Ten minutes of writing, and nothing in it will go stale until the prices change.

How our checker reads your llms.txt

• It requests /llms.txt and /llms-full.txt at the root of the host and does not follow redirects, so a file that only exists after a redirect counts as missing.
• It rejects responses that are HTML, which is what most CMS 404 pages and catch-all routes return.
• It expects a plain-text or Markdown response with real content, not an empty file.
• A valid llms.txt adds its points to the AI Infrastructure part of the score; the file's wording does not change the score.

What each template starts you with

Templates fill the summary line and the sections with placeholder URLs on yourdomain.com. Replace every URL and note before you publish; a file of placeholder links is worse than no file.

Choosing the pages for each section

A useful test for every link: would a person who has never heard of you learn something essential from this page? If not, it belongs in Optional or nowhere.

llms.txt compared with robots.txt and sitemaps

Writing a file worth reading

• Link your clearest, most stable pages: product overviews, pricing, help articles and policies.
• Write link titles and notes so a model understands each page without opening it.
• Keep it short. Twenty excellent links beat two hundred.
• Update it when key pages move. Broken links are worse than no list.
• Add llms-full.txt, the full text of those pages in one file, only if you can keep it current.

Mistakes we see in published files

• Served as HTML. A CMS that returns its 404 page, or a styled HTML page, at /llms.txt. Our checker rejects HTML, because it is not the format.
• Redirected. The file sits behind a redirect to another address. Our checker does not follow redirects for this file, so a redirected file counts as missing.
• No title line. The file starts with links. The # heading is the one required line.
• Placeholder links. Template URLs left in place, or links to pages that now return 404.
• A copy of the sitemap. Hundreds of URLs with no notes. That is what sitemap.xml is for.
• Claims that are not on the site. The summary should describe what the linked pages say, not marketing that appears nowhere else.

When llms-full.txt is worth it

llms-full.txt puts the full text of your key pages into one Markdown file, so a model can read them without fetching each page. It suits documentation sites, where the content is stable, structured and often read in full. It suits marketing sites less well: the file goes stale the moment a page changes, and a stale copy of your pricing is worse than none.

If you publish one, generate it from the same source as the pages during each build, rather than writing it by hand, and keep it under a size you would be happy for someone to read end to end.

Publishing it, and what it adds to your score

1. Download the file, or copy it into a plain-text file named llms.txt.
2. Upload it to the root of the host, for example https://www.example.com/llms.txt.
3. Open it in a browser to confirm it loads as plain text with HTTP 200.
4. Run the free check linked under the generated file to confirm it is detected and valid.

llms.txt is worth 10 points and llms-full.txt 5 points in the AI Infrastructure part of the score (15 points in total), while crawler access is worth 70. That weighting reflects reality: a perfect llms.txt cannot help a site that blocks the crawlers.