A free llms.txt generator that checks its own work
Paste your address. It reads your pages, writes the file from your own titles and descriptions, and checks every line against the v2 format.
What it reads, what it writes
The generator fetches your home page, robots.txt and sitemap, then up to 40 of your pages, once each, and takes every title and description as you wrote it. No model writes a word, so the same site always gives the same file. The file is one part of an AI-ready website.
Your own emphasis first
Pages your home page links to come first, in the order they appear, then the rest of your sitemap. Files, feeds, translated copies, and sign-in, cart and search pages are set aside.
robots.txt is obeyed
A path closed to crawlers is never requested, and a page marked noindex stays out of the file.
Every section gets a turn
When a sitemap lists more pages than the 40 one run reads, each top-level path takes turns, so a long blog can't push out the pricing page.
The v2 order
Your name as the H1, your home page's description as the summary, then sections named after your own paths. Legal pages go under Optional.
Markdown where you have it
A page that advertises a Markdown version of itself is linked by that address, which is the form agents read best.
Six rules, checked as you type
The draft is yours to edit. Every change is checked again against the six rules of the llms.txt v2 format, and each rule that fails says how to fix it.
One H1 title, first in the file
The file opens with a single
# Nameline. It is the only part the proposal requires.A blockquote summary
A
>line or two under the title, holding the facts an agent needs to make sense of the rest.Links grouped under H2 sections
Every link sits under a
## Sectionheading, so an agent can tell what kind of page it is about to open.Every entry is a list link
Inside a section each line reads
- [name](address), with an optional: noteafter it.Full addresses, each listed once
Links start with https://, so they still resolve when the file is read somewhere else, and no page appears twice.
Secondary pages under Optional
Legal and other skippable pages go under
## Optional, the section an agent may leave out when it is short on space.
Where the file goes
Upload the file so it loads at yoursite.com/llms.txt, served as plain text or Markdown. Then link it: the result gives you a describedby link for your pages and the same link as an HTTP header.
Make sure robots.txt lets AI crawlers in, then run the AI-ready check on the live site. It reads the file, the link and the pages behind it.
Does it work?
Not for Google rankings. Google's AI optimization guide says Search doesn't use llms.txt. For agents the file is a map, and they read it when something points them to it.
Ahrefs studied 137,210 domains in May 2026: 97% of the published files got no requests that month, and AI bots made none for files that didn't exist. That is why the link matters as much as the file. What is llms.txt has the evidence, and how AI agents read websites explains who asks for it.
Fair questions
Is the llms.txt generator free?
Yes, with no signup and no email. One run reads up to 40 pages, and one address can make 5 files in ten minutes and 20 in a day.
How do I add llms.txt to my website?
Upload the file so it loads at yoursite.com/llms.txt, as plain text or Markdown. Then add the describedby link the generator gives you to your pages, or send it as a Link header. Without a link, agents rarely ask for the file.
Will llms.txt help my Google ranking, or get me cited by ChatGPT?
Not Google: Google says Search doesn't use llms.txt, and that the file won't help or hurt your rankings. For agents it is a map, not a guarantee. OpenAI's and Anthropic's crawler pages don't mention the file, and server logs show agents fetch it when something points them to it.
Is llms.txt an official standard?
No. It is a proposal by Jeremy Howard, published in September 2024 and revised as v2 on 10 August 2026. No standards body such as the IETF or the W3C has adopted it.
How often should I regenerate it?
Whenever a page is added, renamed or removed, or a price or an address changes. A generated file is a snapshot of the day you made it.
Does it write llms-full.txt?
No. llms-full.txt isn't part of the llms.txt proposal. It is a convention on some documentation platforms: one file holding the full text of every page. It can help on docs sites that agents read end to end, and a marketing site rarely needs it.
What happens to the pages the generator reads?
Each page is fetched once for its title and description, and the result goes straight to your browser. Nothing is stored or published, and no model reads your pages. A short-lived count of requests per address is kept to stop abuse.
A generated file is a snapshot
It is true on the day you make it. Add a page or change a price, and it is out of date until you run the generator again. Laarpi writes llms.txt on every publish, from the same source as the pages, so the two can't disagree.
- An llms.txt that maps the siteOne plain index agents read first, written from the real pages.
- A Markdown twin of every pageThe same words without the layout, linked from llms.txt.
- Structured data in the server HTMLWhat the site is, who runs it and what it sells, in the first byte.