Add a file at /llms.txt on your domain: a plain Markdown document that summarizes what your site is and links to your most important pages. It's not a ranking signal and no major AI vendor has confirmed they use it for training or retrieval, but it costs almost nothing to add, and it gives any crawler or assistant that does check for it a clean, unambiguous summary instead of having to parse your HTML. We publish one ourselves at flashcloud.com/llms.txt, so this is a description of something we actually run, not a theory.
What llms.txt actually is
It's a proposed convention, not a web standard: a Markdown file at your site root, similar in spirit to robots.txt or sitemap.xml but aimed at language models instead of search crawlers or link-following bots. Where robots.txt tells crawlers what they're allowed to fetch, llms.txt tells them what your site is about and points them at the pages worth reading, in a format that doesn't require parsing navigation menus, JavaScript, or marketing layout to find.
The format is deliberately minimal:
- An H1 with your site or company name
- A one-line blockquote summary
- A few sentences of context if useful
- One or more H2 sections, each a list of links with short descriptions
No schema to validate against, no required fields beyond the H1. That looseness is both the appeal and the limitation: it's easy to write, but there's nothing enforcing that any given crawler actually reads it.
Does it help, and with what
Be honest about what this file is and isn't. It is not confirmed to influence ranking in ChatGPT, Perplexity, Claude, or Google's AI features. It is not a replacement for good on-page content, structured data, or a real sitemap. Treat it as a low-cost hedge: if a crawler or retrieval pipeline does look for it, you've made its job easier; if none currently do, you've lost an afternoon.
Where it plausibly helps most is on sites with a lot of navigational and template noise around the actual content, since it gives a shortcut past the chrome straight to what matters. A useful file links to your highest-value, most citable pages: core product or service pages, and any knowledgebase or documentation content written to directly answer questions. It's a companion to your sitemap, not a replacement for one. Your sitemap.xml should still list everything for search engines; llms.txt is a curated, human-readable shortlist for a different kind of reader.
A minimal template
Start simple. A workable structure looks like this:
# Your Company Name
> One sentence describing what you do and who it's for.
Optional paragraph with a bit more context: what makes the product different, who the typical customer is, anything a reader would need before clicking through.
Then group your links under headings, for example:
- ## Docs - links to your core how-to or reference content, each with a short description after a hyphen
- ## Product - links to your main product or pricing pages
- ## Optional - a convention some sites use for secondary links (blog, changelog) that a reader can skip if they're short on context budget
Keep each link's description to a single clause. The point is a fast scan, not a landing page. If you maintain a knowledgebase, this is a natural place to link your best answer-style articles alongside product pages, the same way you'd link them in a sitemap or a footer.
Keeping it current
Treat it the way you'd treat a sitemap: update it when you add or retire pages that belong in the shortlist, not on a fixed schedule. A stale llms.txt pointing at removed pages is worse than no file at all, since it actively misdirects whatever's reading it. If your site publishes new content regularly, such as a blog or knowledgebase, it's worth a quick pass whenever you ship a batch of new articles rather than trying to keep it in perfect sync with every single publish.
There's no dynamic requirement here: a static file, hand-edited or generated by a small script during your build, is enough. If you're running a CDN or edge caching in front of your site, for example through Cloudflare, treat /llms.txt like any other static asset and make sure a fresh deploy actually invalidates the cached copy so crawlers aren't reading a stale version.
Where this fits with the rest of your site
If your site spans multiple domains or subdomains, for instance a marketing site plus a separate app or docs subdomain added as an addon domain or subdomain, each publicly reachable host can carry its own llms.txt scoped to what lives there. There's no requirement to consolidate everything under one file, and a scoped file is usually more accurate than a single one trying to describe several distinct properties at once.
If you don't have a site to put this on yet, that's a separate problem worth solving first: see getting a free website built for you if you have active hosting or a domain with us and no site up.