Free tool — llms.txt generator
Give answer engines a map of your site.
llms.txt tells ChatGPT, Claude and Perplexity what your site is about and which pages matter. Fill in your pages below — the file builds live, per the llmstxt.org spec.
builder runs in your browser · optional crawl reads only public pages · really free
Too lazy to type? Enter your domain — we read your homepage, sitemap and a few core pages once, extract titles, descriptions and internal links, and prefill the builder below. You edit from there.
Upload the file to the root of your domain — yoursite.com/llms.txt — next to your robots.txt.
llms.txt is step one.
A clean llms.txt makes you easier to read — it doesn't tell you whether AI actually recommends you. Run the free visibility scan and see who gets named for your keywords: you, or your competitors.
[ Run the free AI visibility scan ]One markdown file, read by every answer engine.
HTML is written for browsers: navigation, cookie banners, scripts. When an AI assistant reads your site, all of that is noise around the few pages that matter. llms.txt (proposed at llmstxt.org) fixes that with a curated, plain-markdown map at your domain root — the same idea robots.txt had for crawlers, but for meaning instead of permission.
The format
# Acme Docs
> Acme is a CLI for shipping fast.
## Docs
- [Getting started](https://acme.com/docs/start): Install and first steps
- [API](https://acme.com/docs/api)
## Optional
- [Changelog](https://acme.com/changelog)The rules
- # H1 — your site or project name. The only required part.
- > blockquote — a one-line summary an AI can quote directly.
- ## sections — grouped links, each with a short note on what the page answers.
- ## Optional — links that may be skipped when the model's context is short.
Who already ships one
Developer-facing companies adopted llms.txt first — Anthropic, Cloudflare and many docs platforms publish one for their documentation, and docs generators increasingly emit it by default. It costs you one static file; if answer engines look for it, you're on the map. That asymmetry is the whole argument.
Questions.
What is llms.txt?
llms.txt is a proposed web standard (llmstxt.org): a single markdown file at the root of your site that tells AI systems what your site is about and which pages matter. It starts with an H1 title, an optional one-line summary as a blockquote, and H2 sections containing links with short notes — a curated map instead of forcing the model to guess from raw HTML.
Where do I put the file?
At the root of your domain, next to robots.txt: https://yoursite.com/llms.txt. If you also generate llms-full.txt, it lives at the root too. Both are plain text files — upload them like any static asset.
What is the difference between llms.txt and llms-full.txt?
llms.txt is the compact index: title, summary and curated links. llms-full.txt is the expanded, everything-in-one-file variant — one block per page. This generator produces a structured skeleton for it (title, source URL, your note per page); paste the actual page content into each block, or keep the skeleton as a richer index.
Does the Optional section do anything?
Yes — it is part of the spec. Links in a section named Optional may be skipped by AI tools when they need a shorter context. Put secondary pages there (changelog, legal, archive) and keep your core docs in regular sections.
Is my data uploaded anywhere?
The builder itself runs 100 % in your browser — nothing you type leaves the page. If you use the optional "crawl & prefill", we fetch your public homepage, sitemap and a handful of core pages once, server-side, to extract titles and links; the result is cached for one hour and nothing else is stored. No signup either way.
How does "crawl & prefill" work?
You enter a domain, we fetch its homepage and sitemap.xml, collect up to 30 internal links and group them into sections by path — /docs to Docs, /blog to Blog, company pages to an Optional section. We then read up to 20 of those pages (hard limits on size and time) and use each page's title tag as the link title and its meta description as the note. Everything is deterministic parsing; no AI reads your site, and you can edit every field afterwards.