llms.txt template and checklist

What llms.txt is and what it is not

The idea is simple. A language model that visits your site has a limited window of text to work with. Your home page is full of navigation, banners and scripts. An llms.txt file hands over a clean, curated map: who you are, and which pages matter.

Be realistic about the status. Google has said it does not use the file for Search, and no major search engine has told site owners that it ranks or cites pages because of it. The clearest use today is documentation sites, where developer tools and agents can be pointed at the file to fetch a clean index. For most business sites it is a low-cost tidy-up with an uncertain return.

Our view: write it in an hour, publish it, and spend the rest of your time on the pages it links to. If those pages are thin, no file will rescue them.

The template, part by part

The file is Markdown. Every part below is optional except the first line, but the order matters because parsers read from the top.

A filled-in example

Take a small skincare store with one hero product, similar in shape to the Shopify store we built for Belvra. This is a hypothetical file, not something we published for that client. The point is what gets included and what gets left out.

Pre-launch checklist

Each item has our note. A copy of this checklist is meant to be downloadable from this page, so you can tick items off as you go.

• The file loads at your domain root as /llms.txt with a 200 status. Our note: no redirect, no login, no cookie wall.
• It is served as plain text or Markdown in UTF-8. Our note: if your server sends it as a download, fix the content type.
• The first line is a single title and the summary is under 50 words. Our note: read it aloud; if it sounds like an advert, rewrite it.
• Every listed URL returns 200 and is indexable. Our note: a link to a redirected or noindex page tells a model the file is stale.
• Every listed URL is the canonical version. Our note: use the address in the page's canonical tag, with the same trailing slash rule.
• Each line describes what the page contains in one sentence. Our note: 'Prices, delivery times and returns for UK orders' beats 'Our amazing shipping'.
• Nothing private is listed. Our note: the file is public, so staging links and internal tools stay out.
• robots.txt does not block the file or the pages in it. Our note: check both, they conflict more often than you would think.
• A named person owns updates. Our note: add a calendar reminder for every quarter and after each site redesign.

Publish and test

1. Draft the listStart from your five to fifteen most useful pages, not from the sitemap. Our llms.txt generator can give you a first draft to edit.
2. Upload to the rootOn WordPress and Shopify this usually means a plugin, a redirect-free static file or a theme template. On a custom site, drop the file into the public folder.
3. Fetch it as a strangerOpen the address in a private window and with a command-line fetch. Confirm the status, content type and that no cookie banner appears.
4. Watch your server logsLook for requests to the file from AI crawlers over the next month. That tells you more than any claim online about who reads it.

For how the file differs from crawl rules, read llms.txt vs robots.txt.

Mistakes we see most

• Pasting the entire sitemap. A list of 4,000 URLs is not a guide.
• Using the file to stuff keywords. It reads as spam to a person and gives a model nothing.
• Listing pages that have since moved. Dead links make the whole file look unmaintained.