What llms.txt is
llms.txt was proposed by Jeremy Howard of Answer.AI on 3 September 2024. The idea is simple: put a short Markdown file at /llms.txt that says what the site is and links to the pages an AI agent should read, ideally in clean Markdown. The proposal was revised to version 2 in August 2026.
Two things the proposal says about itself are worth knowing before you build one:
- It is a proposal. The spec calls itself a proposal that is open for community input, not a standard.
- It is mainly for inference, not training. The authors expected agents to read it on demand when a user asks for help, though training runs could use it too.
It is also not a replacement for robots.txt or sitemap.xml. robots.txt controls what crawlers may fetch; llms.txt only suggests what to read. A sitemap lists every URL; llms.txt is a short, curated reading list.
Who reads it
| Who | What they officially say |
|---|---|
| Google Search | Its AI optimization guide says Google Search ignores llms.txt. Creating one is fine for other services, but it neither helps nor harms visibility or rankings in Google Search. |
| Chrome Lighthouse | An experimental “Agentic Browsing” audit checks for /llms.txt at the site root. A missing file counts as not applicable; a server error fails the audit. |
| OpenAI, Anthropic, Perplexity | Their crawler documentation covers robots.txt only. None of them says its crawlers read other sites’ llms.txt. All three publish llms.txt files for their own documentation. |
So the honest answer to “does llms.txt help me show up in AI answers?” is: nobody has promised it does. It mostly helps coding agents and other tools that look for it on purpose, which is why it is most common on software documentation sites.
The format (v2)
The file is Markdown, with parts in a fixed order:
| Part | Required? | What goes in it |
|---|---|---|
| Byte-order mark | No | Allowed at the very start |
| H1 | Yes | The name of the site or project. This is the only required part |
| Blockquote | No | A short summary with the key facts needed to understand the rest |
| Paragraphs or lists | No | Any extra context, but no headings |
| H2 sections | No | Lists of links: - [Name](URL), optionally followed by : notes |
A section named ## Optional is, by convention, for secondary links an agent can skip when it needs a shorter context. In version 1 that section had a special meaning to tools; version 2 dropped it, so it is now only a convention.
A copyable example
# Example Bakery
> Example Bakery is a family bakery in Springfield that sells bread and cakes online and delivers within 20 km.
All prices are in US dollars and include tax. Orders placed before 2 pm ship the same day.
## Shop
- [Bread](https://example.com/bread.md): Daily loaves, ingredients and allergens
- [Cakes](https://example.com/cakes.md): Custom cakes, sizes and lead times
## Policies
- [Delivery](https://example.com/delivery.md): Areas, fees and time slots
- [Refunds](https://example.com/refunds.md)
## Optional
- [Our story](https://example.com/about.md)Replace the names and links with your own. Keep the summary factual: the same claims should appear on your real pages.
Where to put it
- At the site root,
https://yoursite.com/llms.txt. Version 2 also allows files in subpaths such as/docs/llms.txt; a file covers the pages under its path, and the most specific one applies. - If you only make one, make the root one: Lighthouse only looks at the root.
- Serve it as plain text with a 200 status, and keep it small enough to fit in an AI model’s context. Put detail in the linked pages, not in the file.
Markdown versions of pages (optional)
The proposal also suggests offering a clean Markdown copy of important pages at the same URL, either with .md added (page.html.md) or with the extension replaced (page.md). Version 2 recommends linking them with rel="alternate" type="text/markdown", and pointing to the covering llms.txt with rel="describedby", in an HTML <link> or an HTTP Link header. This is a recommendation, not a requirement.
What llms-full.txt is
llms-full.txt is not part of the proposal. It is a convention from documentation platforms such as Mintlify: the whole documentation site merged into one file. It can be useful for very large docs, but nothing in the llms.txt spec asks for it.
Should you add one?
Yes, if your site has pages that explain what you sell, how it works and what it costs: it takes a few minutes and costs nothing. But do it after the basics, because llms.txt cannot fix them:
- Make sure AI crawlers are allowed in
robots.txtand not blocked by your CDN. See AI crawlers: the list, and how to allow or block them. - Make sure your content is in the HTML, not only rendered by JavaScript.
- Then add a short, accurate llms.txt and keep it updated when pages change.
Citeloop’s own file is at citeloop.ai/llms.txt. The free GEO audit, opening soon, will check whether yours exists and follows the format.