llms.txt Generator
Build an llms.txt file: a short Markdown document you place at the root of your domain that tells language models what your site is, what it covers and which pages are worth reading. Fill in the fields and copy the result.
- Free, no sign-up
- Runs in your browser
- Nothing uploaded
- Updated Sep 2026
At a glance
- What it is
- A Markdown file at
/llms.txtsummarising a site for AI systems - Format
- H1 title, blockquote summary, then H2 sections of links
- Status
- A community proposal, not an official standard — adoption varies
- Replaces
- Nothing. It sits alongside robots.txt and your sitemap
- Effort
- Ten minutes for most sites
- Processing
- Entirely client-side
What llms.txt is, and what it is not
llms.txt is a proposal for a single Markdown file at the root of a domain
that describes, in plain language, what the site contains and which pages best represent
it. The reasoning behind it is straightforward: a language model working from a web page
has to wade through navigation, cookie banners, related-post widgets and advertising to
reach the content. A curated file removes that work, and makes it more likely the model
reads what you would want it to read.
The structure is deliberately simple. An H1 with the site name, a blockquote giving a one-paragraph summary, optionally a few lines of context, then H2 sections containing Markdown links with short descriptions. A section titled Optional has special meaning — it marks links that can be skipped when context is limited.
Two things it is not. It is not an access-control mechanism: it does not grant or deny permission to crawl, and it has no bearing on whether your content is used for training. Those questions belong in robots.txt and your terms of service. And it is not a ranking factor. No search engine has stated that it influences rankings, and anyone claiming otherwise is guessing.
Is it worth publishing?
An honest assessment, because there is a good deal of overclaiming about this file.
The case for: it costs very little to produce, it cannot hurt, and a handful of AI tools and documentation platforms do read it. If your site is documentation, an API reference or a tool catalogue — content people ask assistants about — the file gives you some influence over which pages get surfaced. Several large developer-documentation sites publish one.
The case against: adoption is genuinely limited. The major AI crawlers have not committed to reading it, and a file nobody fetches produces no benefit. Treat any claim that it will increase your AI visibility as unproven.
The sensible position: publish one if your content is the kind people ask assistants about, treat it as cheap insurance rather than a strategy, and put your real effort into the things that demonstrably affect how AI systems represent your content — clear page structure, direct answers near the top of the page, factual accuracy, and structured data that states plainly what each page is.
Writing a file that is actually useful
The common mistake is listing every page. A file with four hundred links is no more useful to a model than a sitemap, and the whole point is curation.
Include the pages that define what your site is: the main documentation entry points,
your most substantial guides, the tools or products people come for, and an about page that
establishes who you are. Twenty to fifty links suits most sites. Write a genuine
description after each link — [Token counter](/ai-token-counter): Estimates token
counts for GPT, Claude and Gemini without sending text to a server tells a model
something; [Token counter](/ai-token-counter) does not.
Leave out login pages, cart and checkout flows, tag and category archives, paginated listings, and anything thin. Use absolute URLs, keep the summary blockquote to a couple of sentences that would make sense to someone who has never heard of you, and update the file when the site changes materially — a stale map is worse than none.
How to use the llms.txt Generator
-
Enter your site name and summary
The summary becomes the blockquote at the top. Two sentences explaining what the site is and who it is for, written so it makes sense with no other context.
-
Add your key pages
Group them into sections — Documentation, Tools, Guides. Give each link a one-line description of what is on that page.
-
Mark anything skippable as Optional
Links in a section named Optional are understood to be lower priority and safe to skip when context is tight.
-
Save it at your domain root
Upload the file so it is reachable at
https://yourdomain.com/llms.txt. It must be at the root, exactly like robots.txt.
Frequently asked questions
Does llms.txt improve my Google rankings?
There is no evidence that it does, and Google has not said it reads the file. Treat it as a way to give AI systems a cleaner view of your content, not as an SEO tactic. Anyone selling it as a ranking factor is speculating.
Is it an official standard?
No. It is a community proposal that some sites and tools have adopted. There is no governing body and no guarantee any particular crawler reads it. That is the main argument for keeping the effort you spend on it proportionate.
Does it stop AI companies training on my content?
No. It has no access-control meaning whatsoever. To restrict AI crawlers, use robots.txt with the relevant user-agent tokens, and state your position in your terms of service. Even then, robots.txt is a request that well-behaved crawlers honour, not an enforcement mechanism.
Where exactly does the file go?
At the root of your domain, served as text/plain or text/markdown from https://yourdomain.com/llms.txt. A file in a subdirectory will not be found.
How is it different from a sitemap?
A sitemap is an exhaustive machine-readable list of every URL, meant for search crawlers deciding what to fetch. llms.txt is a short curated summary in prose and links, meant for a model deciding what to read. Publish both — they answer different questions.
How often should I update it?
Whenever the site changes materially — a new product area, a restructured documentation tree, a section retired. A file pointing at pages that no longer exist actively misleads, which is worse than not having one.