llms.txt Generator A sitemap lists everything. This says what matters.
Build a draft
- Up to 30
- Your browser
- Nothing
Point it at a domain and it reads the sitemap, then the pages, then writes the file — every line built from that page’s own title and description. What comes back is a draft, and how good it is tells you something.
What the file is for
The format is three moves: name the thing, say what it is in one line, then list the pages that explain it, grouped and described.
# Acme
> Acme builds payroll software for teams of 10 to 200.
## Start here
- [Home](https://acme.io/): What Acme does and who it is for.
- [Pricing](https://acme.io/pricing): Three plans, billed per employee.
## Documentation
- [API reference](https://acme.io/docs/api): Endpoints, auth, and rate limits.The syntax is the easy half and the generator handles it. The half that carries the value is deciding which pages belong and what each one is really for — which is an argument about a business, and not something a crawler can make on your behalf.
Publishing it is cheap and reversible. No engine is obliged to read it, and the honest position is that adoption is still uneven. The reason to do it anyway is that answering the question costs an afternoon and improves the pages themselves.
Questions
What is llms.txt?
A single markdown file at the root of a site — /llms.txt — that lists the pages worth reading and says what each one is for. It is a proposal rather than a ratified standard, and no engine is required to honour it. It is cheap to publish and it costs nothing if ignored, which is roughly the whole argument for doing it.
How is it different from a sitemap?
A sitemap is an inventory: every URL that exists, with no opinion about any of them. An llms.txt is an argument: these are the pages that explain us, in this order, and here is what each is for. The first is for a crawler with unlimited budget. The second is for a model with room to read twenty pages and not two thousand.
Will publishing llms.txt get me cited?
On its own, no, and anyone promising otherwise is selling something. Adoption is uneven and no major engine has committed to reading it. What it does reliably is force the editorial question — which twenty pages actually explain this business — and that question is worth answering whether or not any crawler ever reads the file.
Why do the descriptions in my generated file look repetitive?
Because they are your meta descriptions, unedited. That is the point of generating rather than writing from scratch: the draft shows you exactly what a model sees when it reads your pages. If thirty pages come back describing themselves the same way, the file is not the problem it found.
Should I publish the draft as-is?
No. Read it, cut what does not belong, reorder what does, and rewrite the lines that say nothing. The generator does the mechanical part — finding the pages and formatting the file. The judgment about what matters is the part that carries the value, and it is not automatable.
A curated map is worth nothing to an engine that was never allowed to fetch the pages on it. That is the first question, and it is the one worth going back and confirming. Check which crawlers can reach you