Free tool
llms.txt generator.
Build a valid llms.txt: a title, a one line summary, and curated sections of links that tell a language model what your site is and which pages actually matter. Switch to full mode and it builds llms-full.txt too, with the content inlined rather than linked. It is a proposed convention rather than a standard anybody has committed to, so this page is honest about what it is worth. Everything runs in your browser and nothing is uploaded.
Your site
The name and summary become the heading and the blockquote at the top of the file. Both are worth getting right: they are the part most likely to be read.
Your sections
One link per line, as Title | URL | note. The note is optional, and so is the third column. Paste straight from a spreadsheet if you have one.
The content itself
One box per link above. Your browser cannot fetch your own pages for you, so this is paste or type. If you have more than a handful of pages, generate the file from whatever already holds the content rather than doing it here.
Add a name to start.
Worth knowing first
What this file is, and what it is not.
llms.txt is a proposal: a markdown file at your root that gives a language model a curated index of your site instead of leaving it to guess from your navigation. The format is deliberately simple, which is most of its appeal.
Nobody has committed to reading it. Google has said it does not use the file. No other major provider has confirmed that it does. The convention may be adopted or may quietly disappear, and anyone claiming it gets you cited in AI answers is inventing a mechanism that has not been demonstrated.
It is still probably worth publishing. It is a static text file. It costs one afternoon and no ongoing effort, it cannot hurt your rankings, and if adoption happens you are already there. That is the entire argument, and it is a reasonable one as long as nobody sells it to you as more.
The real work is upstream of the file. If a model describes your business wrongly, the cause is almost always what third party sources say about you rather than what your own root directory contains. That is covered on LLM SEO, and it is where the effort actually belongs.
The format
What a valid file looks like.
The structure is short enough to hold in your head. An H1 with the name, a blockquote summary, optional prose, then H2 sections each containing a markdown list of links. A section named Optional is understood to be skippable when context is short.
# Acme Analytics > Self hosted product analytics for small teams, without cookies. Runs on your own infrastructure. No third party scripts and no personal data leaves your servers. ## Core pages - [How it works](https://example.com/how-it-works): the architecture in plain terms - [Pricing](https://example.com/pricing): per server, published ## Guides - [Self hosting on a single box](https://example.com/guides/single-box) ## Optional - [Changelog](https://example.com/changelog)
FAQ
Questions about llms.txt
Does anything actually read llms.txt?
No major provider has confirmed that it does, and Google has said plainly that it does not use it. Anthropic, OpenAI and Perplexity have not committed either way. Anyone telling you a file at your root gets you cited in ChatGPT is selling something.
So the honest case for publishing one is not that it works today. It is that it costs almost nothing, it is a static text file with no downside, and if the convention is adopted you already have it. We publish llms.txt and llms-full.txt on this domain on exactly that reasoning, and you can fetch both and check them.
What is the difference between llms.txt and llms-full.txt?
llms.txt is an index: a title, a short summary, and curated links grouped into sections, so a model can see the shape of your site and fetch what it needs. It is meant to stay small.
llms-full.txt is the expanded version, with the actual content inlined rather than linked, for a model that wants everything in one request. This tool builds both: the toggle above the output switches between them, and in full mode you get a content box per link.
One honest limit on the full version. A browser cannot fetch your pages to fill those boxes for you, because the same origin policy blocks it, so the content is paste or type. That is fine for a dozen pages and miserable for two hundred. Past that, generate the file from whatever already holds the content: this site builds its own llms-full.txt from the same catalogue that feeds its sitemaps, which is the pattern worth copying.
Is this the same as robots.txt?
No, and the two are often confused. robots.txt controls what a crawler is permitted to fetch and is enforced by convention on the crawler's side. llms.txt does not permit or forbid anything. It is a curated summary offered to whoever wants it.
They also solve opposite problems. If your concern is AI crawlers taking your content, llms.txt is not the lever and adding one changes nothing. That is a robots.txt and firewall decision.
Which pages should go in it?
The ones you would send a knowledgeable stranger to, which is usually far fewer than you would list in a sitemap. An index that repeats your whole URL set is not an index, it is a sitemap in a worse format.
Put the pages that explain what you do and answer real questions. Put the pages you would be happy to be quoted from. Leave out tag archives, thin category pages and anything you would not want representing you out of context.
Does the Optional section do anything?
It is part of the proposed spec, and it means what it says: a model short on context can skip that section without losing the point of your site. Treat it as a place for secondary material rather than a dumping ground, and if you would mind it being ignored entirely, it does not belong there.
Keep reading
Related
Book me
Want to know why a model describes you wrongly?
Send the business name and what you sell. You get a straight answer on what the models currently say, where they are getting it from, and whether any of it is fixable.