What llms.txt is and why you’re hearing about it in 2026
llms.txt is a proposed standard from Jeremy Howard, published in 2024 — a Markdown file at the root of your domain that tells AI tools and their search systems which pages on your site matter, and gives each one a short description. It’s a young idea: not a formal standard, but a community-maintained spec. As of 2026, Claude and Perplexity declare support, while Google, based on public statements, doesn’t use it. So for most sites the question isn’t “how do I write one” — it’s whether to bother with it at all right now.
This guide covers who supports llms.txt in 2026 and who doesn’t, how to write and place it correctly, when it’s worth deploying, and when your time is better spent elsewhere. It’s different from the guide to robots.txt for AI crawlers, which covers AI crawler access rather than navigation for AI.
The state of support in 2026
Adoption of llms.txt is uneven as of mid-2026. The picture below is based on Anthropic’s documentation, Perplexity’s announcements and documentation, public statements from Gary Illyes and John Mueller, and the April 2026 changes to Lighthouse.
llms.txt support across AI platforms
- Anthropic Claude declares llms.txt support in its documentation (since January 2026); the file can act as a navigation aid for picking out important pages.
- Perplexity declares support (since mid-2025); it can use the file to orient itself among your site's important URLs.
- ChatGPT (OpenAI) hasn't officially confirmed support; if you test llms.txt there, treat the results as rough observations only.
- Google per a July 2025 statement, doesn't support it and has no plans to; it did add an llms.txt check to Lighthouse in April 2026 as a quality check.
The key takeaway: if Google is your primary channel, don’t expect llms.txt to pay off directly in classic search or in AI Overviews. If you’re targeting Claude or Perplexity users (typically B2B, documentation, developer tools), llms.txt can be worth it, because those tools declare support for it.
llms.txt vs robots.txt
These two files are easy to mix up, but they do completely different jobs:
| Aspekt | robots.txt | llms.txt |
|---|---|---|
| Purpose | Controls crawler access to pages (allow/disallow) | Points AI to important pages with a brief description |
| Age and standard | A standard since 1994, widely respected | Proposed in 2024, community spec, uneven support |
| Location | Root of the domain (/robots.txt) | Root of the domain (/llms.txt) |
| Who respects it | Mainstream search engines typically do; with AI crawlers it depends on the operator | Mainly Claude and Perplexity; Google doesn't |
Short version: robots.txt is a must-have, llms.txt is an optional add-on depending on what you’re targeting.
The llms.txt format: how to write one
Jeremy Howard’s spec is simple — Markdown with headings and links:
# Site name
> A short description of what the site does.
## Key pages
- [Home](https://yourdomain.com/): the main entry point
- [Services](https://yourdomain.com/services/): an overview of what we offer
- [Pricing](https://yourdomain.com/pricing/): current pricing
## Documentation
- [Guides](https://yourdomain.com/docs/): step-by-step walkthroughs
- [FAQ](https://yourdomain.com/faq/): the most common questionsYou can optionally add an llms-full.txt with the complete text content of your key pages in Markdown — useful mainly for documentation, when you want to hand AI the full content directly instead of making it load individual URLs.
When to deploy llms.txt and when not to
The decision comes down to the type of site:
- Deploy — documentation, libraries, developer tools, complex B2B SaaS sites (multiple content types), and sites whose audience uses Claude or Perplexity to research.
- Consider — mid-sized sites with mixed content (an e-commerce site with a blog plus documentation), if you have time to maintain it.
- Skip for now — small brochure sites (1–3 pages), a strategy aimed only at Google, and sites with no capacity for regular upkeep.
A poorly maintained llms.txt with stale links or links to 404s can point the tools that do support it toward worse entry points. If you don’t have the capacity to keep it current, you’re better off not deploying it yet.
Common mistakes and myths
Treating llms.txt as a ranking signal
No direct effect in Google; for Claude and Perplexity it’s navigation, not ranking.
Fix: It isn't a signal for ranking results.
Deploying it on a small site with no maintenance plan
On a 1–3 page site it adds no value; once it goes stale, it just gets in the way.
Fix: Think about whether it's worth it.
Confusing llms.txt with robots.txt
robots.txt controls access, llms.txt points to content.
Fix: They're two different files with different purposes.
Putting it anywhere but the root
The proposed spec says AI should look for it at the root; anywhere else, you can’t count on it being found.
Fix: Always /llms.txt at the root of the domain.
What’s next: put llms.txt in context
In 2026, llms.txt is a supplementary layer for AI search with uneven support. On a typical company site, quality content, structured data, reputation, and managing AI crawler access through robots.txt all come first. For the full picture of AI visibility, see the practical SEO for AI checklist.
And if you’re weighing whether llms.txt makes sense for your specific site, judge it as part of a broader AI visibility strategy — not in isolation.
