What is an llms.txt file?
An llms.txt file is a Markdown file at /llms.txt that gives AI assistants and agents a short, curated map of your site. A coding agent looking up your API, or a chat assistant answering a question about your product, can read it first and then open only the pages it needs.
Jeremy Howard proposed the format in September 2024, and the llmstxt.org spec was last updated in August 2026. It sits next to the files you already have, and does a different job from each:
| File | Who reads it | What it does |
|---|---|---|
| llms.txt | AI assistants and agents, on demand | A curated summary of your key pages, with a note on each link |
| robots.txt | Crawlers, before they fetch | Says which paths each crawler may fetch |
| sitemap.xml | Search engines | Lists every indexable URL, with no explanation |
How to generate an llms.txt file
Enter your domain and the generator builds a first draft from your live site in under a minute. The steps:
- Type your website address, like acme.com, and choose Generate llms.txt.
- The tool fetches your home page, robots.txt and sitemaps, then reads up to 30 pages: the ones your home page links to first, then a sample spread across your sitemap's sections.
- Read the draft. Fix titles, cut links you don't want agents to start from, and rewrite the summary line if it says TODO.
- Copy or download the file and publish it at the root of your domain as plain text.
- Switch to Validate and check the live URL to confirm your server returns the file, not an HTML page.
What the generated llms.txt contains
The draft follows the spec's order: an H1 with your name, a blockquote summary, then H2 sections of links. Your name comes from the Organization markup or og:site_name on your home page, else your title tags. The summary is your home page's meta description.
Each page is typed by its URL and title, then grouped into Product, Use cases, Comparisons, Integrations, Docs, Blog and resources, Pages and Company. Each link's note is the page's meta description, or its first sentence when there is none. Legal pages and big programmatic sections (more than eight pages under one path) go to the Optional section, which agents may skip when context is short.
Pages marked noindex, pages that point their canonical elsewhere and pages that fail to load are left out. This is the same builder LogNorm uses when it generates llms.txt for a workspace, where a model types every page and the full crawl covers far more of your site.
llms.txt format and example
The only required part is the H1. Everything else is optional but useful, in this order:
# Acme
> Acme is project-management software for small teams: plans, tasks and status updates in one place.
Acme runs in the browser and has iOS and Android apps.
## Product
- [Pricing](https://acme.com/pricing): Plans, limits and what each includes
- [Integrations](https://acme.com/integrations): Slack, GitHub and Google Drive
## Docs
- [Getting started](https://acme.com/docs/start): Create a workspace and your first project
## Optional
- [Changelog](https://acme.com/changelog): Release notes by monthParagraphs between the summary and the first H2 can explain anything an agent needs to read the links correctly, but they can't contain headings. Inside an H2 section, every item is a list item that starts with a [name](url) link, optionally followed by a colon and a note.
How the llms.txt validator checks a file
The validator reads a live URL or pasted text and sorts what it finds into three levels:
- Errors break the spec: no H1, more than one H1, content above the H1, headings in the details part, or list items that don't start with a link.
- Warnings are allowed but weaken the file: no summary, empty sections, duplicate links, most links without notes, a leftover TODO, or a file over 50,000 characters.
- Notes are suggestions: relative URLs, no Optional section, or most links pointing to other sites.
For a live file, it also shows the content type your server sends and flags an HTML response. LogNorm's own GEO audit uses a five-point quality score built from the same checks.
Does llms.txt help SEO or AI visibility?
llms.txt doesn't move Google rankings. On 15 June 2026 Google added a note to its AI optimization guide saying these files aren't needed for Google Search, won't help or hurt your visibility, and are fine to keep for other services that use them.
Those other services are the reason to have one. According to llmstxt.org, thousands of sites publish the file, documentation platforms generate it, and Chrome's Lighthouse checks for one in its agentic browsing audits. Treat it as a ten-minute job that helps agents read you correctly, not as a ranking lever. For being mentioned in AI answers, the content on your pages matters far more: see GEO vs SEO.
Keep llms.txt current with LogNorm
A hand-written llms.txt goes stale the week you ship a new page. In LogNorm, the GEO audit checks whether /llms.txt and /llms-full.txt exist, scores the file and lists its problems. From the audit you can generate both files from your crawled pages, and your coding agent can commit them to your repo. The next GEO audit confirms what the live file now serves.