llms.txt: What It Is, Who Reads It and How to Write One
llms.txt is a Markdown file at the root of a website (/llms.txt) that gives AI tools a short map of the site: a title, a one-line summary, and lists of links to the pages that matter most, each with a note. Jeremy Howard proposed it in September 2024 at llmstxt.org. Google says it doesn’t use the file for Search, AI Overviews or AI Mode, so treat it as a cheap courtesy to AI agents, not a ranking factor.
That last point is where most of the confusion sits. Plenty of articles sell llms.txt as the robots.txt of AI search. It isn’t, and knowing what it does and doesn’t do saves you from spending a week on it.
What is llms.txt, exactly?
A plain Markdown file with a fixed shape. The proposal defines it in this order:
- An H1 with the site or project name. The only required part.
- A blockquote with a short summary of what the site is.
- Optional paragraphs or lists with anything a reader needs to make sense of the links.
- H2 sections of links, each a Markdown list item:
- [Page name](https://example.com/page): what's on it. - An “Optional” H2 section by convention, for secondary links an agent can skip when it’s short on context.
The proposal also suggests a clean Markdown version of each page at the same URL with .md added, so an agent can read the content without the menus, scripts and cookie banners. Some sites publish an llms-full.txt as well, with every page’s text in one file; that’s a common convention, not part of the proposal.
Who actually reads llms.txt?
Fewer readers than the hype suggests. Here’s where things stand as of October 2026:
| Reader | Does it use llms.txt? | What that means for you |
|---|---|---|
| Google Search, AI Overviews, AI Mode | No. Google’s Search Central guidance (May 2026) says you can ignore it | Rankings and AI Overview citations still come from your HTML |
| Chrome Lighthouse | Its Agentic Browsing audit checks for the file | A missing file can show up as a flag in your audit |
| AI coding agents reading developer docs | Often, when a docs site publishes one | Docs sites and APIs get the most out of it |
| AI labs’ own docs | The proposal lists OpenAI, Anthropic and Gemini as publishers of llms.txt for their developer docs | They publish it; that doesn’t mean their crawlers use it to rank you |
| SEO tools | Many check for it | Expect it in audit reports |
Google’s John Mueller called the file “purely speculative for now” in June 2026 and said he knew of no crawlers or LLMs using it beyond SEO tools. He also said it may help developer documentation with AI coding agents. That matches the table: real value for docs, little for a typical marketing site.
Should you add an llms.txt to your site?
Add one if any of these is true:
- You publish developer docs, an API or an MCP server that people connect to from an AI tool.
- Your audits flag it and you want a clean report.
- You’d like AI agents that do read it to land on your best pages first.
Skip it, or leave it for last, if you’re hoping it will get you cited in ChatGPT or Google’s AI answers. The work that moves citations is on the page itself: a direct answer in the first lines, question headings that match how people ask, sourced facts and AI crawlers allowed in robots.txt. The generative engine optimization guide covers that work, and the FAQ section guide covers the part that’s easiest to get right.
How do you write an llms.txt file?
Twenty minutes, if you already know which pages matter.
- Write the H1 and the summary. The name of the site, then one sentence that says what it is and who it’s for. Plain words, no slogan.
- Pick 10 to 30 pages. Your homepage, pricing, the docs or setup page, and the posts that answer your buyers’ main questions. Not every URL: an agent should be able to read the whole file in one go.
- Group them under H2s. “Start here”, “Docs”, “Guides”, “Optional”. One line per link, with a note that says what’s on the page, not “Learn more”.
- Save it as
/llms.txtat the root, served as plain text. - Keep it current. If your site has a blog, generate the file from the same source as your sitemap so new posts appear without anyone editing it.
A minimal file looks like this:
# Example Co
> Example Co makes invoicing software for freelance designers.
## Start here
- [Pricing](https://example.com/pricing): plans, prices and the free trial
- [Setup guide](https://example.com/docs/setup): connect your bank in five steps
## Optional
- [Changelog](https://example.com/changelog): every release since 2024
We do this on copix.si: our llms.txt is built from the site’s config and the blog, so each new post shows up on deploy, and /llms-full.txt adds every post in full.
What does llms.txt not do?
It doesn’t block or allow anything. Crawler access lives in robots.txt, and an AI crawler that isn’t allowed there won’t fetch your pages whatever llms.txt says. It doesn’t add pages to an index, it doesn’t replace a sitemap, and nothing guarantees an AI assistant will read it before answering a question about you.
It also can’t fix a page that doesn’t answer the question. If a post buries the answer under 300 words of intro, a link to it in llms.txt sends agents to the same weak page. The blog post structure guide and the guide to whether AI content hurts SEO cover the page side.
How do you check your llms.txt and the rest of your AI readiness?
Open https://yoursite.com/llms.txt in a browser: it should load as text, start with an H1 and every link should work. Then check robots.txt for the AI crawlers you want to let in.
If you use Copix Complete, the SEO Engine does both in one call. Ask your AI to audit a page and it runs the page rules, scores the page’s AI-citation readiness, reads robots.txt for search and AI crawlers and checks llms.txt on the host, for 3 credits. A crawl of 25 pages (10 credits) probes llms.txt along with broken links, redirects and orphan pages. To see whether ChatGPT, Claude, Gemini and Perplexity actually cite you, the GEO Engine asks them a query with web search on and shows who they cite instead. The website SEO checker guide walks through a full audit.
What does Copix cost?
Copix is an MCP server: it works inside the AI you already use (Claude, ChatGPT, Cursor or Claude Code). Copix Basic is $9 a month for write, review and your saved brief, unlimited. Copix Complete is $29 a month or $290 a year and adds the SEO, GEO, Conversion and Ads engines with 1,500 credits every month, rolling over up to three months. Every paid call tells you its price in credits and waits for your yes.
Your first plan starts with a 14-day free trial: card at checkout, nothing charged for 14 days, then $29 a month for Complete unless you cancel from your account.
Start your 14-day free trial of Copix Complete, or see what each engine does first.
FAQ
Is llms.txt a ranking factor?
Google’s Search Central guidance says the file plays no part in its rankings or its AI features, and other AI search products haven’t confirmed using it either.
What is the difference between llms.txt and robots.txt?
robots.txt tells crawlers what they may fetch, and well-behaved crawlers follow it. llms.txt is a reading list: it suggests which pages matter and what’s on them, and controls nothing.
Do I need llms-full.txt too?
You can skip it. It’s an extra some sites publish with the full text of every page, and it helps agents that want the whole site in one request, and it’s easy to generate if your pages are already Markdown.
Can I write llms.txt with ChatGPT or Claude?
A model can draft it well when you give it your real page list and what each page is for. Check every link before you publish: a model will happily invent a /docs/api page you don’t have.
Where does llms.txt go?
It sits at the root, https://yoursite.com/llms.txt. The proposal also allows one under a subpath, such as /docs/llms.txt, covering the pages below it.