llms.txt Generator: Create, Import and Check Your File

This free llms.txt generator builds the file that shows language models and AI agents which pages on your site matter. It follows the current specification v2 from llmstxt.org, can import your sitemap or an existing llms.txt, and checks every line. Below the tool you will find which pages belong in the file and the mistakes I see most often. What the file is, and what it cannot do, I explain in What Is llms.txt?

Go to the generator

llms.txt Generator Built on Specification v2

How to create an llms.txt in four steps

With the generator above, your llms.txt file comes together like this:

1. Enter the name of your company or project. It becomes the H1, the only required part of the file. The short description below it says in one or two sentences what you offer and to whom.
2. Add sections such as services, pricing and documentation, and enter the pages. If you have a sitemap, paste it under "Import". The generator sorts the URLs into sections by their first path segment.
3. Write a short note for each link saying what the page is for. Agents use that note to decide which link to follow.
4. Read the check on the right, then copy the file or save it as llms.txt.

Where you can, link to a Markdown version of the page. Specification v2 places it at the same URL with a .md ending, such as /pricing.md or /pricing.html.md. An agent reads the content there without navigation, scripts or cookie banners.

Which pages belong in your llms.txt

Your llms.txt should hold the pages an agent needs to answer a question about your offer: services or products, pricing, documentation, contacts. According to the specification, the file gives a curated overview and stays small enough to fit in a language model's context. The complete list of all pages belongs in sitemap.xml, not in llms.txt. How the two files differ is covered in llms.txt vs. robots.txt vs. sitemap.xml.

If you import a sitemap, trim it afterwards. Archive pages, tags and single blog posts rarely help an agent. Useful but secondary links go into the "Optional" section at the end. Since v2 that section is only a convention: tools no longer drop its links automatically, but an agent may skip them.

You decide the order of sections. My own llms.txt lists services first, then tools and research, and company and legal pages last. Each section names the English page before the German one.

The most common llms.txt mistakes

The generator found the first mistake on my own site. The llms.txt of litzki-systems.com was built from the same text as the machine view of the home page. 17 of 22 sections contained prose and lists without links, 178 lines in total. I rebuilt it on Oct 29, 2026. Since then my build checks the file on every change with the same rules as the generator.

1. Prose inside H2 sections. Under v2 these sections hold link lists only, in the form - [Title](URL): note. Explanatory text goes between the blockquote and the first H2.
2. Missing blank lines. A paragraph that follows the blockquote without a blank line becomes part of the quote in Markdown. The generator inserts the blank lines for you.
3. Relative URLs such as /pricing/. The specification does not require full URLs, but its examples use them throughout, and a copied file stays unambiguous. Enter your domain and the generator completes them.
4. URLs with http:// or a redirect. Every fetch then takes a detour. Link the final https URL.
5. Outdated links. An llms.txt that points to moved or deleted pages sends agents nowhere. Regenerate the file whenever a linked page changes, ideally automatically from your CMS or build.

Where to put your llms.txt and how to check it

Save the file as UTF-8 under exactly this name in your root directory, for example your-domain.com/llms.txt. Since v2 it may also live in a subpath such as /docs/llms.txt. It then covers every URL below that path; if several files apply, the most specific one wins. The specification does not prescribe a content type. text/plain and text/markdown are common.

Then open the file in your browser. In Lighthouse, Google only checks whether /llms.txt responds without a server error. A missing file with a 404 counts as not applicable, a server error as a finding. New in v2 is a pointer from every page to the file that covers it, as <link rel="describedby"> in the page head or as an HTTP header. The generator shows the code under "Make it discoverable".

This check does not tell you whether the file has any effect. For that, the content behind the links has to be right and match your structured data. How to check that is covered in how to validate your llms.txt.

Frequently asked questions

Is the llms.txt generator free?

Yes, with no sign-up. The tool runs entirely in your browser, sends nothing to a server and sets no cookies. What you enter is gone when you close the page.

Why doesn't the generator crawl my website?

Because a good llms.txt is a selection, not a list of every page. Generators that crawl a site take over many pages automatically, and some write the descriptions with AI. Here you paste your sitemap, keep the pages that matter and write the notes yourself. Your data never leaves the browser.

Do I also need an llms-full.txt?

No, it is not part of the specification. An llms-full.txt that bundles the text of every page is a common addition that some documentation platforms generate. I offer one on litzki-systems.com and link it in the "Optional" section.

How do I add an llms.txt in WordPress or Wix?

According to the list on llmstxt.org, Yoast SEO and All in One SEO for WordPress, as well as Wix, have their own llms.txt feature. Use one route only: either that feature or the file from the generator, uploaded to the root directory via FTP or a file manager.

llms.txt, JSON-LD and agentic SEO in one review

I review the llms.txt, structured data and access rules of your site together and show what an agent actually finds there. This is part of Agentic SEO and GEO. The initial call takes thirty minutes and is free, with no obligation. You get every finding in writing afterwards.

Book an initial call ↗ Prefer to write