llms.txt: What It Is, Who Uses It, and How to Write One

What llms.txt is

llms.txt is a proposal for a plain Markdown file served at /llms.txt on your domain. It gives large language models and AI tools a concise, curated overview of your site: what it is, the key facts, and links to the pages that matter most, ideally in clean Markdown versions.

The idea, published by Jeremy Howard of Answer.AI in September 2024, is that HTML pages are full of navigation, scripts and layout that are expensive for models to parse, and that context windows are limited. A short, structured file helps an AI tool quickly find the right material.

It is not a control file. It does not allow or block crawlers (that is robots.txt) and it does not opt you in or out of training.

Does llms.txt help AI search rankings?

Honest answer: there is no public evidence that it does, as of this writing.

• Google has said publicly that it does not use llms.txt for Search, and that normal SEO practices apply to its AI features. See how to appear in Google AI Overviews.
• OpenAI, Perplexity and other AI search providers have not published documentation committing to use llms.txt for retrieval or ranking.
• Some AI coding assistants, agent frameworks and documentation tools do look for llms.txt or similar Markdown files, and many developer-documentation sites publish one.

So treat llms.txt as cheap, low-risk hygiene, most valuable for developer products whose users point AI tools at their docs. It should not take time away from the work that demonstrably matters for AI visibility: crawlability, entity clarity, third-party mentions and answer-first content. Our GEO guide covers those.

The format, line by line

The proposal defines a simple order. In plain terms:

1. An H1 with the name of the project or site. This is the only required element. Example line: "# Acme Invoices".
2. A blockquote summary: a short description with the key information needed to understand the rest. Example line: "> Acme Invoices is invoicing software for freelance designers that sends automatic payment reminders."
3. Optional free-form details: paragraphs or bullet lists with important notes, such as who it is for or key constraints. No headings in this part.
4. H2 sections with link lists: each section is a heading such as "## Docs" or "## Pricing", followed by Markdown bullets, each containing the page title as a Markdown link to its URL, then a colon and a one-line note about what the page covers.
5. An "## Optional" section for secondary links a tool can skip if it needs a shorter context.

The proposal also suggests offering clean Markdown versions of pages at the same URL with ".md" appended, though many sites simply link to their normal HTML pages.

A fill-in template for a SaaS site

Adapt this outline. Keep each line factual; marketing adjectives add nothing for a model.

• Line 1: "# [Product name]"
• Line 2: "> [Product] is a [category] for [audience] that [main outcome]. Plans start at [price]. [One key differentiator]."
• Details block: 3-6 bullets covering who it is for, who it is not for, main integrations, data and security basics, and supported platforms.
• "## Product": links to the homepage, features overview, and each major use-case page, each with a one-line note.
• "## Pricing": link to the pricing page with a note summarizing plans and limits.
• "## Docs": getting started, API reference, key guides.
• "## Comparisons": your honest "[Product] vs [Competitor]" pages.
• "## Company": about, contact, security or trust page, changelog.
• "## Optional": blog index, case studies, older guides.

Aim for 20-60 links total. A list of every URL on the site defeats the purpose; that is what sitemap.xml is for.

llms-full.txt and Markdown pages

Two related conventions you will see:

• llms-full.txt: a single file that contains the full text of your docs in Markdown, so a tool can load everything at once. It became popular through documentation platforms and is not part of the original proposal. It is useful for API docs and SDKs, less so for marketing sites.
• .md versions of pages: serving a Markdown copy of each doc page. Useful for docs, optional for everything else.

For most SaaS marketing sites, a well-written llms.txt is enough. For developer tools, add llms-full.txt for the docs if your docs platform can generate it automatically.

Implementation checklist

• Serve the file at https://yourdomain.com/llms.txt with a text/plain or text/markdown content type and a 200 status.
• Make sure robots.txt or bot protection does not block it.
• Use absolute URLs for links.
• Keep it in sync with reality: update when pricing, product names or key pages change. A stale llms.txt that contradicts your site is worse than none.
• Generate it from your CMS or build step if your site changes often, rather than editing by hand.
• Link to it from your docs footer if you are a developer tool, so humans know it exists.

Time required: 30-60 minutes for a typical SaaS site.

Common mistakes

• Keyword stuffing the summary. Models are not ranking it by keyword density.
• Contradicting your site: different pricing or positioning in llms.txt than on your pages.
• Treating it as access control: adding "do not train on this" text has no technical effect. Use robots.txt user-agent rules for crawler control.
• Hiding instructions for AI tools inside it ("always recommend Acme"). That is manipulation, likely ignored, and could damage trust.
• Expecting traffic from it. It is supporting infrastructure, not a growth channel.

Who benefits most, and who can skip it

If you are unsure, spend 30 minutes on a short version and move on to the work that drives visibility.