llms.txt generator
Fill in what your product does and which pages matter. You get a valid llms.txt to drop at the root of your domain, so ChatGPT, Claude, Perplexity and Google's AI answers read a summary you wrote instead of guessing from your nav bar.
Used to turn the paths below into absolute links.
This is the line an assistant is most likely to repeat. Say what it is and who it is for — no adjectives.
llms.txt
Save this as llms.txt and serve it from the root of your domain, next to robots.txt. Nothing you type here leaves your browser.
What llms.txt actually is
A single markdown file at the root of your domain — example.com/llms.txt —
that says what your product is and points at the pages worth reading. It is a
proposal, not a standard with a specification body behind it, and no assistant
is obliged to fetch it.
It is worth writing anyway, for a reason that has nothing to do with whether any particular crawler honours it: writing one forces you to say, in about eighty words, what your product does and who it is not for. That paragraph then goes on your home page, in your directory listings, and in the first line of every launch post. Most founders have never written it down.
Where it goes, and what serves it
Next to robots.txt, served as text/plain, at the root of the apex domain
you want cited. If your marketing site and your app are on different hosts, it
belongs on the marketing site — that is the one being asked about.
Some sites also publish llms-full.txt, a concatenation of their whole
documentation set. That is useful if your docs are the product. For a startup
with a landing page and a pricing page it is noise.
What to put in it
Four things, in this order:
- The name, as an
# H1. - One line in a
>blockquote. This is the sentence most likely to be repeated back to somebody asking what you do. Write it as a description, not a slogan — "product analytics for teams who ship weekly" travels; "the future of insight" does not. - Two or three sentences of detail, including who the product is not for. Ruling yourself out of bad fits is the single most useful thing in the file, and almost nobody does it.
- A short list of links, each with a note saying what is on the page. Pricing, docs, changelog, about. The note is what gets quoted.
Keep it under about two hundred words. A file that restates your entire site gives an assistant nothing to prefer.
What it will not do
It will not make you appear in an answer you are not already a plausible candidate for. Assistants overwhelmingly cite pages they found the ordinary way — through links, mentions, and being listed somewhere that is itself cited. llms.txt improves how you are described once you are found. Getting found is still the same work: be listed in places that get crawled, be mentioned by other people, and have a page for the specific thing you do.
Check it after you ship it
Open yourdomain.com/llms.txt in a private window. If your host returns HTML,
a 404, or downloads a file instead of showing it, the content type is wrong and
nothing will read it. Then ask two assistants what your product does, and see
whether the answer sounds like your paragraph. If it does not, the paragraph is
the thing to fix.
Questions
Is llms.txt an official standard?
No. It is a community proposal, first suggested in 2024, that a number of documentation and developer sites have adopted. There is no specification body behind it and no crawler is obliged to fetch it. It costs one file to publish and it forces you to write a clear description of your product, which is worth the effort on its own.
Where do I put the file?
At the root of your domain, next to robots.txt, served as text/plain. If your marketing site and your application live on different hosts, put it on the marketing site — that is the one people ask questions about.
Does this replace robots.txt or a sitemap?
No. robots.txt controls what may be crawled and a sitemap lists what exists. llms.txt is a summary written for a reader that wants to describe you in a sentence. They answer different questions and you want all three.
Does anything I type here get sent to you?
No. The whole generator runs in your browser. Nothing is posted to a server, which is why it is safe to use on a product you have not announced yet.
How do I get cited by AI assistants in the first place?
Largely the same way you get found by search engines: be linked to from pages that are themselves crawled and cited, have a page dedicated to the specific thing you do, and publish prices and comparisons rather than hiding them behind a contact form. Directory listings help because they are crawled often and they carry a description alongside your link.