# llms.txt: what it is and when it is worth creating

> What an llms.txt file contains according to its specification, what it doesn't do for your rankings, and a table to decide whether your site needs one.


**llms.txt is a Markdown file a website publishes at `/llms.txt` to tell agents and language models what the site contains and where its most useful information lives.** Jeremy Howard proposed it in September 2024, and the [specification](https://llmstxt.org/) is now in its second version, from August 2026. It is an open proposal: no search engine requires it to crawl, index or show a page.

## What an llms.txt file contains

The specification asks for a fixed structure, so the file can be read both by a language model and by a program:

1. An H1 title with the name of the site or project. It is the only required part.
2. A blockquote (`>`) with a short summary and whatever is needed to understand the rest of the file.
3. Optional paragraphs or lists with context, without further headings.
4. H2 sections with lists of links in the format `[name](url): notes`.

By convention, a section called `Optional` holds secondary links an agent can skip when it needs to save context. The second version adds two ideas: publish a Markdown version of each page at the same URL ending in `.md`, and declare it in the HTML with `rel="alternate" type="text/markdown"`. It also allows an llms.txt in a subfolder, such as `/docs/llms.txt`, covering only that part of the site.

Here is what one could look like for a fictional pottery studio:

```markdown
# Luz Pottery Studio

> Studio in Valparaíso with beginner and intermediate classes,
> custom tableware for restaurants and an in-store shop.

Classes are booked online and custom orders are quoted by email.
Prices are in Chilean pesos, VAT included.

## Classes
- [Beginner pottery](https://example.com/classes/beginner.md): eight sessions, materials included, dates and spots
- [Intermediate wheel](https://example.com/classes/wheel.md): requirements, schedule and price

## Custom orders
- [Tableware for restaurants](https://example.com/orders.md): lead times, minimum order and how to get a quote

## Optional
- [Studio history](https://example.com/history.md)
```

The file is short on purpose. The detail lives on the linked pages, and the agent opens them only when it needs them to answer.

## What it doesn't do

- **It doesn't control access.** That is what [robots.txt](/en/guides/robots-txt-explained/) is for. The specification itself separates the two roles: robots.txt says what automated access is considered acceptable; llms.txt is read when an agent needs information to help someone.
- **It doesn't replace the sitemap.** A [sitemap](/en/guides/what-is-a-sitemap/) lists URLs so search engines can discover them; llms.txt picks and explains a few.
- **It isn't a requirement to appear in Google.** Google says that for its AI features, such as AI Overviews and AI Mode, [you don't need to create new machine-readable files, AI text files or special markup](https://developers.google.com/search/docs/appearance/ai-features).
- **It doesn't guarantee anyone reads it.** Each agent decides whether to check it. According to its author, the heaviest use is in software documentation, where coding agents follow it to find API references.

## Who uses it today

The specification states that thousands of sites publish one, that documentation platforms generate it automatically, and that AI labs offer one for their developer docs. For example, the [Claude documentation](https://platform.claude.com/llms.txt) publishes one.

At HeyMark we generate [our llms.txt](https://heymark.ai/llms.txt) from the same content the website publishes: product, pricing, MCP server tools and guides. That way there is no third copy of our prices that someone has to remember to update. Each article also has a Markdown version at the same URL ending in `.md`.

## Does your site need one?

| Type of site | Decision | Why |
| --- | --- | --- |
| Documentation for a product, an API or an MCP server | Create it | It is the most common use: agents and assistants read that documentation to use the product |
| Small services website | Optional, if it's cheap | It can summarize services, prices and policies in one place; don't expect an effect in Google |
| Store or catalog | After the basics | First make every product page crawlable and indexable; then an llms.txt can point to shipping and returns |
| Blog or publication with hundreds of articles | Only if it's generated automatically | By hand it goes stale quickly; link to indexes or sections instead of every article |

## How to decide with your own data

Before investing in maintaining it, publish it and check your server logs: how many requests `/llms.txt` gets and from which user agents. If nobody asks for it in weeks, there's no need to expand it. If agents that matter to your audience read it, keep it as current as a pricing page: a broken link or an old price in llms.txt gives the agent a wrong answer.

If what you want is for your content to appear and be cited in AI search answers, the main work is elsewhere: in letting their crawlers in and in pages that answer clearly. We explain it in [generative engine optimization (GEO)](/en/guides/generative-engine-optimization/).

## Plan your next post in HeyMark.

Keep the idea, review the draft with your team, and see how it performed in the accounts you connected.

[Start free](https://app.heymark.ai) · [See how it works](https://heymark.ai/en/#product)

## Developer and agent resources

- [HeyMark MCP documentation](https://heymark.ai/en/mcp/)
- [llms.txt](https://heymark.ai/llms.txt)
- [Full site content for language models](https://heymark.ai/llms-full.txt)
- MCP protocol endpoint: `POST https://mcp.heymark.ai`
- OAuth protected-resource metadata: [/.well-known/oauth-protected-resource](https://mcp.heymark.ai/.well-known/oauth-protected-resource)
- MCP server card: [/server-card](https://mcp.heymark.ai/server-card)
