SecondDoor

llms.txt: worth writing, not worth believing in

What belongs in an llms.txt, and the evidence on whether anything reads it. Cheap insurance, not a strategy.

What the file is for

An llms.txt is a plain text index of your site, written for a reader that cannot execute JavaScript and has no interest in your layout. It sits at the root, at /llms.txt, and it is served as text/plain.

A sitemap tells a crawler which URLs exist. An llms.txt tells an assistant which of them matter and what each one is. That difference is the whole point: the assistant is choosing what to read, not what to index.

The shape that works

Start with an H1 naming the company. Follow it with a single blockquote paragraph saying what you do and who for. That paragraph is the part an assistant quotes when someone asks about you, so it is worth more attention than the rest of the file combined.

Then group your links under H2 headings by section: documentation, pricing, product, support. Each link gets the URL and a short note on what is there. Keep it under a few hundred lines. It is an index, not a mirror.

The common mistakes

Serving HTML at /llms.txt is the most frequent one. A single page app that answers every unknown path with its shell will return 200 and a page of markup, and every checker will read that as a missing file.

The second is writing marketing copy in the summary paragraph. An assistant summarising your product will reuse your words. Vague words in, vague words out.

Questions

Does llms.txt replace robots.txt?
No. robots.txt controls whether a crawler may read your site. llms.txt helps a reader that is already allowed find the right pages. You need both, and they answer different questions.
Will an assistant definitely read it?
Probably not. Google stated in June 2026 that llms.txt has no effect on Search or AI Overviews, and server log studies show the major assistant crawlers rarely fetch it. Adoption sits somewhere between 0.2% and 10% depending on whose crawl you read. Write it because it costs an afternoon and cannot hurt, not because it will move anything on its own.
So what should I do instead?
The two things the evidence supports: do not block the assistant crawlers, and serve your content in raw HTML rather than assembling it client side. Those gate everything else, and most sites failing them are failing by accident.
Where should it live?
At the root of the domain, at /llms.txt, served as text/plain. Not in a subdirectory, and not behind a redirect chain.
How machine discoverability is scoredCheck your own site
llms.txt: worth writing, not worth believing in, SecondDoor