On this page
What is llms.txt?
llms.txt is a plain Markdown file served at https://yoursite.com/llms.txt. It tells AI assistants, in a few lines, what your website is and which pages matter, so a model does not have to guess from menus, cookie banners and scripts.
Jeremy Howard of Answer.AI proposed it in September 2024, and it is free to publish.
Here is the start of the one on this website:

llms.txt, robots.txt and sitemap.xml: the difference
All three sit at the root of your site and are read by machines, but they answer different questions.
| File | Question it answers | Who reads it | Format |
|---|---|---|---|
robots.txt |
Who may crawl which pages? | Every crawler, including AI crawlers | Plain text rules |
sitemap.xml |
Which pages exist? | Search engines | XML, every page |
llms.txt |
Which pages are worth reading, and what is this site? | AI assistants and agents that look for it | Markdown, a short list |
They work together. robots.txt still decides access: listing a page in llms.txt does not unblock it. See AI crawlers and robots.txt for which bots to allow.
Does llms.txt help SEO or AI search?
Be clear about this before you spend time on it:
- Google: no. Google’s own guidance says you do not need “new machine readable files, AI text files, or markup” to appear in AI Overviews or AI Mode. Good pages are what count.
- AI assistants: some AI tools and coding agents fetch llms.txt when it exists. None of the big assistants has said it uses the file to choose what to quote.
- So why publish one? It takes minutes, it costs nothing, and writing it forces a useful question: which 20 pages explain your site best? Those are also the pages you want AI answers to quote.
We keep one live on every website we manage, and MonoRanks checks each week that it is still there and still valid.
The format, line by line
The proposal at llmstxt.org defines the structure. Only the first line is required.
# Acme Bikes
> Mountain, gravel and city bikes, sized and shipped from Portland.
We build and sell bikes online. Every bike ships fully assembled.
## Guides
- [Mountain bike sizes](https://acmebikes.example/guides/sizes): how to pick a frame size
- [Tubeless setup](https://acmebikes.example/guides/tubeless): step by step, 20 minutes
## Products
- [Trail 29](https://acmebikes.example/bikes/trail-29): hardtail, 12-speed
## Optional
- [Company history](https://acmebikes.example/about/history): founded 2014
| Part | Required? | What to write |
|---|---|---|
# Name |
Yes | The site or company name, once |
> Summary |
No, but always add it | One sentence that says what the site is for |
| Plain text | No | A few lines of context: who it is for, what is special |
## Section with links |
No | Groups of pages, each as - [Title](URL): what it covers |
## Optional |
No | Pages an AI can skip when it needs a short version |
The optional parts, with examples
llms.txt has no fixed list of fields like robots.txt has. Everything after the heading is ordinary Markdown, so the “extra parameters” you see on other sites are conventions: plain lines that help a reader. These are the ones worth using:
1. Context lines under the summary. Short facts a model would otherwise have to dig for:
# Acme Bikes
> Mountain, gravel and city bikes, sized and shipped from Portland.
- Website: https://acmebikes.example
- Ships to: United States and Canada
- Contact: hello@acmebikes.example
- Last updated: 2026-10-07
2. Link descriptions. Every link can carry a description after a colon. This is the most useful part of the file:
- [Tubeless setup](https://acmebikes.example/guides/tubeless): step by step, 20 minutes, no special tools
3. The Optional section. The proposal reserves the heading ## Optional for pages an AI can skip when it needs a short version: history, press, old guides.
4. Markdown copies of pages. The proposal also suggests offering a clean Markdown version of important pages at the same address plus .md, for example https://acmebikes.example/guides/sizes.md. Documentation sites do this; most business sites skip it.
Things that are not part of llms.txt: Allow or Disallow rules (that is robots.txt), keyword lists, and tags such as <meta>. A model reads the file as text, so anything that is not useful text is noise.
Websites in more than one language
The proposal does not cover languages, and AI tools look for the file at the root only. What works in practice:
- One file at the root, in your main language, at
https://yoursite.com/llms.txt. - A Languages section that lists the home page and the most important pages of every language, so a model knows the other versions exist.
- Optional: a file per language at
/de/llms.txt,/es/llms.txt, linked from the root file. Nothing reads those on its own, so the link from the root is what matters.
# Acme Bikes
> Mountain, gravel and city bikes, sized and shipped from Portland.
## Guides
- [Mountain bike sizes](https://acmebikes.example/guides/sizes): how to pick a frame size
## Languages
- [Deutsch](https://acmebikes.example/de/): German version of the site; [Größenberatung](https://acmebikes.example/de/ratgeber/groessen): Rahmengröße wählen
- [Español](https://acmebikes.example/es/): Spanish version; [Tallas](https://acmebikes.example/es/guias/tallas): cómo elegir la talla
This website uses the same pattern: one English file at the root, ending with a Languages section that points to the German, Spanish and Persian pages.
Rules of thumb
- Describe every link. The few words after the colon are the part a model reads first.
- Use full URLs (with
https://), plain Markdown, no HTML. - Group by the way a visitor thinks: Guides, Products, Pricing, Company.
- Keep it short: 10 to 30 links. A 400-line file tells a model nothing.
How to create your llms.txt
With the free generator (any website, two minutes)
Paste your domain into the MonoRanks llms.txt generator. It reads your home page and up to fourteen linked pages and drafts an llms.txt in the standard format. Check the summary line, remove pages that do not matter, and publish it.

On WordPress
You probably already have a plugin that does it. The four big SEO plugins all create llms.txt in their free versions: Yoast SEO, Rank Math, All in One SEO and SEOPress. Turn it on in the plugin’s settings and check the live file.
- Check what it lists. Plugins often list your latest posts. Swap in your key pages if the setting allows it.
MonoRanks drafts the file from your own pages and, with its WordPress plugin, publishes it after you approve it.

On Astro and other static sites
Add the file to your public folder (public/llms.txt) so it is served at the root. Better: generate it at build time from your pages, so it never goes out of date. This website does that with a small endpoint, src/pages/llms.txt.ts, that lists every page with its title and description.
On hosted builders
Shopify, Wix, Squarespace and similar platforms do not all let you place a file at the root. Check your platform’s help pages or app store for “llms.txt”; some need an app or a redirect.
How to check it works
- Open
https://yoursite.com/llms.txtin a browser. You should see plain text, not your 404 page. - The first line starts with
#and the second with>. - Click three or four links. Each should open a live page, not a redirect or a 404.
- Check that robots.txt does not block the AI crawlers you want; llms.txt cannot override it. The free AI crawler check shows both in one go.
What not to put in it
- Marketing copy or keyword lists. Write it like an index, not an ad.
- Every page you have. That is what the sitemap is for.
- Private, draft or outdated pages.
- Links that redirect. Use the final address.
llms-full.txt: do you need it?
Some sites publish a second file, llms-full.txt, with the full text of their main pages in one Markdown file, so an AI tool can read everything in one request. It is a common convention, not part of the llms.txt proposal.
For documentation and developer sites it is useful. For most business websites, a good llms.txt is enough.







