llms.txt Generator

Write a correct llms.txt in the browser — site name, one-line summary, and the pages you want read first, in the order you want them read.

This becomes the H1, and it is the one line the proposal actually requires.
Rendered as a blockquote under the H1. Write it as a sentence a stranger could repeat back: what you are, who for, what makes you different.
Free prose. Use it for the things a reader would otherwise get wrong: what your terms mean, what is out of scope, which figures are current.
A line starting with # opens a section. Every other line is Title | URL | Description. A section named Optional is the one special case in the proposal — it marks pages a model may skip when it is short of room.
llms-full.txt is the same site in one long file with the actual page content inlined. This tool cannot fetch your pages, so it builds the skeleton and you paste the markdown in.
Your file 

          

          

            
          

          

          
Nothing you type here is uploaded, stored or sent anywhere. It all runs in your browser.

        

      

    

  


  

    

      
What llms.txt is, and what it is not

      
llms.txt is a convention proposed by Jeremy Howard in September 2024: a markdown file at the root of your domain that tells a language model what the site is and which pages to read. The format is small on purpose — an H1 with the site name, a blockquote summary, optional prose, then H2 sections of markdown links, each with a short description after a colon.

      
Here is the part most pages selling you on it leave out. It is a proposal, not a standard. It has real adoption among documentation sites — several large developer-docs platforms generate one automatically — and as of this writing no major AI provider has publicly confirmed that they read it. Google has said publicly that it does not use it. So treat any claim that llms.txt gets you cited as marketing, including if you read it on a page like this one.

      
The honest case for publishing one is that it costs an afternoon, it does nothing bad, and writing it forces you to answer "which ten pages actually matter here" in a way most sites never have. If adoption arrives you are ready; if it does not you have lost very little.


      
The file that matters more, and already works

      
Everything in llms.txt is a hint that something might one day read. Your robots.txt is read today, by every crawler that respects it, and it decides whether an answer engine can see the page at all. If you are spending effort on being legible to AI, spend it there first — the AI Crawler Checker shows you bot by bot what you are currently allowing, and it is common to discover you have been blocking the crawler that produces citations while allowing the one that only trains.


      
How to choose what goes in it

      
    
        
  • Write descriptions for a machine with no time. The description is what decides whether a link gets opened. "Pricing" says nothing; "Retainer tiers, what is included at each, and the minimum term" says when to open it.
    • 
        
  • Order is the signal. There is no ranking in the file other than the order you write. Put the page you would show a stranger first, first.
    • 
        
  • Use Optional for the nice-to-haves. It is the only section name with defined meaning: content a model may drop when context is tight. Your About page belongs there. Your pricing does not.
    • 
        
  • Link to clean pages. A URL in llms.txt that redirects, or that renders its content only in JavaScript, wastes the one thing the file was for.
    • 
      


      
Where it goes

      
Save it as llms.txt at the root of the domain, served as text/plain or text/markdown, so it resolves at https://yourdomain.com/llms.txt. There is no need to reference it from anywhere else; the convention is that it is found by its location. If you also publish llms-full.txt, that goes beside it.

    

  


  

    

      
Questions people ask

      

        

          
Does ChatGPT or Claude actually read llms.txt?

          
No provider has publicly confirmed that they do, and Google has said it does not. Anyone telling you otherwise is guessing. Publish it because it is cheap and because writing it clarifies your own site, not because you have been promised a result.

        

        

          
Is llms.txt a replacement for a sitemap?

          
No. A sitemap is a machine-readable list of every indexable URL, for crawlers, and it is a real standard. llms.txt is a short curated reading list with descriptions, for a model with limited context. Keep the sitemap.

        

        

          
What is llms-full.txt?

          
The same idea with the content inlined rather than linked: one long markdown file containing the actual text of the pages, so a model can read the whole site in a single fetch. It suits documentation and small sites. On a large site it becomes a file nobody maintains, which is worse than not having one.

        

        

          
Does publishing llms.txt let AI companies train on my site?

          
It has no bearing on that either way. Training access is governed by your robots.txt and by the licence of your content. If you want to allow some crawlers and block others, generate the rules with the AI Crawler Control Generator — llms.txt is not a permission file.

        

        

          
Do I have to keep it up to date?

          
If it drifts it is worse than absent, because it points a reader at pages that no longer say what you told it they say. Treat it like the footer navigation: small enough to be reviewed whenever the site structure changes, and not longer than you are willing to maintain.

        

      

    

  



  

    

      
Related tools

      
    

  


  

    

      
A file that lists your best pages is only as good as the pages

      
llms.txt points at your work. Whether that work is worth pointing at, and whether anything else on the web agrees, is the part I help with.

      

        Book a Call
        More free tools