How to Create an llms.txt File

Knowing how to create an llms.txt file takes about thirty minutes and five steps. An llms.txt file is a plain Markdown page at the root of a site that gives AI agents a short, curated map of the pages worth reading. Choose the pages, write a summary, group the links, save the file at /llms.txt and test it. OpenAI, Anthropic and Google publish one for their developer docs. Whether any AI system reads yours is a separate question, and the dated sources below answer it.

What is an llms.txt file?

The llms.txt proposal asks websites to publish a Markdown file at /llms.txt that offers brief background, guidance and links to the detailed pages a language model should read. Jeremy Howard first published it on 3 September 2024, and the current page is a second version, last modified on 10 August 2026.

The file can sit at the site root or at any path inside it. A file at /docs/llms.txt covers the pages under /docs/, and where more than one file applies, agents are expected to use the most specific one.

The idea behind it is simple. A web page is built for people, so its useful text is wrapped in navigation, scripts and ads. A model that has to answer a question wants the useful text and a short list of where to find more. An llms.txt file is that list, written by the person who knows the site best.

What does an llms.txt file contain?

The format is Markdown with a fixed order. Only the first item is required.

1. An H1 with the name of the site or project.
2. A blockquote with a short summary that holds the key information needed to understand the rest.
3. Optional paragraphs or lists of any kind, except headings, with more detail.
4. Optional H2 sections, each holding a list of links. Every item is a Markdown link, optionally followed by a colon and a note.
5. An H2 section named Optional, by convention for secondary links that an agent can skip when it needs a shorter context.

The example below is a made-up shop, Acme Supply, written to that format.

# Acme Supply

> Acme Supply sells reusable packaging to small food businesses. This file lists the pages needed to answer questions about products, prices and shipping.

Prices are in US dollars. Orders ship within two business days.

## Key pages

- [Pricing](https://acme-supply.example/pricing.md): Plans and volume discounts
- [Product catalog](https://acme-supply.example/products.md): Every container, lid and bag
- [Shipping and returns](https://acme-supply.example/shipping.md): Delivery times and return rules

## Optional

- [About Acme Supply](https://acme-supply.example/about.md)
- [Packaging guides](https://acme-supply.example/blog.md): Guides on food packaging rules

Real files follow the same shape at larger scale. The OpenAI API docs file opens with an H1, a one-line blockquote and a note that each entry has a Markdown twin. Stripe’s file opens with a blockquote about the company and then lists pages under headings such as Payments. Cloudflare’s file links to one llms.txt file per product.

How to create an llms.txt file in five steps

Step 1: Choose the pages

List the pages that answer the questions a customer or an agent asks most often. For a shop that means pricing, the product catalog, shipping, returns and contact. For a software product it means the docs home, the quick start, the API reference and the changelog.

The proposal says the file should stay small enough to fit in context, with the detail behind the links. It gives no page count. A list of 10 to 25 links is a workable range for most sites, and a short file beats a complete one.

Step 2: Write the title and the summary

Use the site name as the H1. Write the summary as a blockquote of one or two sentences that state what the business does, who it serves and where. Use plain words, because the proposal asks for concise, clear language and for jargon to be explained or left out.

A good summary can be read on its own. If an agent stopped after the blockquote, it should still know what the site is.

Step 3: Group the links under H2 headings

Split the links into a few sections with names a reader would recognize, such as Docs, Products or Policies. Write each item as a link followed by a short note that says what the page contains. The note matters more than the title, because it tells an agent whether to open the page.

Put secondary pages under an H2 named Optional. The proposal reserves that name for links an agent can skip when it is short on space.

Step 4: Point each link at a clean page

The proposal recommends a clean Markdown copy of each page at the same URL with .md added, for example /pricing.md. OpenAI’s file takes that route. Stripe’s file links to its normal HTML pages instead, which also works.

If Markdown copies do not exist yet, link the normal pages. If they do exist, link those. The markdown versions check explains how to produce them.

Step 5: Save the file and test it

Save the file as llms.txt in the root of the site so that https://yourdomain.com/llms.txt returns it. Documentation can have its own file at /docs/llms.txt.

Then run the test the proposal suggests: ask an AI agent questions about your site while giving it only the llms.txt file as a starting point. If it cannot find the pricing page, the notes need work. Also confirm the file loads without an error. Chrome’s Lighthouse reports a server error on the file, and treats a plain 404 as not applicable.

Writing the file by hand is fine, and several tools generate one. The proposal lists the WordPress plugins Yoast SEO and AIOSEO, Wix, the documentation platforms Mintlify and GitBook, and plugins for Docusaurus and VitePress. The free llms.txt generator from LLMReady builds the file in the browser from a list of pages.

Who reads an llms.txt file today?

The answer depends on whom you ask, and the sources disagree in tone. This table lists what each one says, with its date. Every page was opened on 1 November 2026.

Three readings follow from the table.

• It is not a Google lever. Google’s own pages say AI Overviews and AI Mode need no extra AI files, so an llms.txt file will not change how a page appears there.
• Its real use is on demand. The proposal describes the file as information used when an agent needs a topic while helping a user, and says it is used most heavily for software documentation, where coding agents follow it to API references.
• No source promises citations. Nothing opened for this guide says an llms.txt file makes an assistant cite or recommend a site.

Is an llms.txt file worth creating?

For most sites it is a small, cheap job with a modest and uncertain return. Build one when any of these is true:

• The site has documentation or an API that coding agents are likely to read.
• Product, pricing or policy pages are scattered across menus and hard to find.
• Clean Markdown copies of key pages already exist or are planned.
• You want a record of which pages you consider the source of truth.

Skip the file as a ranking tactic. The same thirty minutes spent on crawler access, a correct sitemap and pages that load their text without scripts will do more for how machines read a site.

How is an llms.txt file different from robots.txt and a sitemap?

The three files do different jobs, and none replaces another.

The proposal makes the same split. A sitemap often lacks the machine-readable versions of pages, leaves out external links and covers far more than fits into a context window. The proposal says an llms.txt file can complement robots.txt by providing context for the content that robots.txt allows, so the crawler rules come first.

Which mistakes make an llms.txt file less useful?

• Listing every page. That is the job of a sitemap. A curated file is the point.
• Vague notes. “Click here” tells an agent nothing. A note such as “Delivery times and return rules” does.
• Stale links. Old URLs that redirect or fail waste an agent’s time. Re-check the list whenever pages move.
• Blocking the bots that should read it. A file that invites agents while robots.txt shuts them out sends two opposite signals.
• Listing private pages. The file is public, like robots.txt. Include only pages you are happy to have read.

Frequently asked questions

Do I need an llms.txt file?

No. Google says AI features need no extra files, and Chrome calls the file optional. It is worth adding when agents are likely to read your documentation or product pages, and the cost is low.

Where does an llms.txt file go?

Put it at the root, so that https://yourdomain.com/llms.txt returns it. The proposal also allows a file at any path, such as /docs/llms.txt, which covers the pages under that path.

Does an llms.txt file allow or block AI crawlers?

Neither. Access rules live in robots.txt. The llms.txt file only describes the content that is available. See the guide to robots.txt rules for AI crawlers for the allow and block lines.

Will an llms.txt file get my site cited by ChatGPT?

No source opened for this guide says so. OpenAI’s crawler page mentions llms.txt only as the index of its own docs. Whether a page can appear in ChatGPT search depends on OAI-SearchBot being allowed in robots.txt, and on the page itself.

Generate your first llms.txt file in two minutes

Open the free llms.txt generator, paste your key pages in the form Title, URL and note, and copy the finished file. Save it at the root of the site and run the test from step 5. Now that you know how to create an llms.txt file, the generator removes the typing. LLMReady also checks that the file loads, that its links resolve and that it sits beside Markdown copies of your pages, as described in the llms.txt check. Join the beta list at LLMReady early access to see that report for your own site.