What Is llms.txt? AI Visibility Guide 2026

llms.txt is a plain markdown file you place at the root of your website (/llms.txt) that gives AI assistants a curated map of your most important content. Instead of letting ChatGPT, Claude or Perplexity guess what matters on your site, you hand them a short index: who you are, what you do, and which pages to read first. Think of it as a sitemap written for language models instead of search engines.

Why llms.txt exists

llms.txt exists because AI assistants have limited context windows, and your website is bigger than what they can actually read. When a model answers a question about your product, it doesn't browse your entire site like a human would. It sees fragments — whatever fits into its context window at that moment.

That creates a real problem for small sites and indie products:

• Your homepage is full of marketing copy, but your docs page is where the actual answers live.
• Your pricing is buried behind a JavaScript-heavy page that some crawlers never render.
• Your best explainer blog post competes with 200 other pages for the model's attention.

Search engines solve this with crawling, indexing and ranking over months. AI assistants often answer in seconds. llms.txt shortcuts the whole process: you decide which pages represent you, and you put that decision somewhere machines are trained to look.

What goes in an llms.txt file

An llms.txt file is valid markdown with three required parts: an H1 with your project or company name, a blockquote that summarizes what you do, and one or more H2 sections that list your key links with short descriptions.

Here is a minimal example for a fictional SaaS:

# Acme Analytics

> Acme Analytics is a privacy-first web analytics tool for indie
> developers. One script tag, no cookies, GDPR compliant by default.

## Docs

- [Getting started](https://acme.example/docs/start): install the
  tracking script in under two minutes
- [API reference](https://acme.example/docs/api): REST endpoints for
  pulling your stats programmatically

## Product

- [Pricing](https://acme.example/pricing): free up to 10k pageviews,
  then flat pricing
- [Comparison with Google Analytics](https://acme.example/vs/ga):
  what we do differently and why

## Company

- [About](https://acme.example/about): who builds Acme and why

A few rules worth following:

• Keep it short. The whole point is to save the model from reading everything. If your llms.txt links to 300 pages, you've rebuilt your sitemap and gained nothing.
• Write descriptions, not just URLs. One sentence per link tells the model when that page is relevant.
• Link to clean, content-rich pages. Docs, guides, pricing and comparisons work well. App dashboards behind logins do not.
• Keep the H1 and blockquote accurate. Models may quote these lines directly when describing your product.

llms.txt vs robots.txt vs sitemap.xml

The three files do different jobs, and llms.txt replaces neither of the others. Here is the honest comparison:

| | robots.txt | sitemap.xml | llms.txt | |---|---|---|---| | Audience | Crawler bots (search + AI) | Search engine crawlers | AI assistants and LLM agents | | Purpose | Say what bots may access | List every indexable page | Point to your most important content | | Format | Plain text rules | XML | Markdown | | Who reads it | Googlebot, Bingbot, GPTBot, ClaudeBot… | Mostly Google and Bing | ChatGPT, Claude, Perplexity and other LLM tools | | Replaces the others? | No | No | No |

robots.txt is a gatekeeper — it controls whether AI crawlers can enter at all. If you're accidentally blocking GPTBot or ClaudeBot there, no llms.txt file will save you. We wrote a full guide on AI crawlers and robots.txt if you want to check yours.

sitemap.xml is exhaustive — it says "here is everything, you decide what matters." llms.txt is opinionated — it says "here is what matters, in my own words." You want all three.

How to create your llms.txt in 10 minutes

You can write a solid llms.txt by hand in one sitting. Here is the process:

1. List your 5–15 most important pages. Homepage, docs quickstart, pricing, your best comparison page, your two or three strongest blog posts. If a page wouldn't help someone decide whether your product fits their problem, skip it.
2. Write your H1 and blockquote. H1 is your product name. The blockquote is 1–3 sentences: what it is, who it's for, and the one thing that makes it different. Write it like you'd want an AI assistant to repeat it — because it might.
3. Group your links under H2 sections. Common groupings: Docs, Product, Blog, Company. Each link gets a title and a one-sentence description.
4. Save it as llms.txt at your site root. It must be reachable at https://yoursite.com/llms.txt, served as plain text. On most static hosts (Cloudflare Pages, Netlify, Vercel) you just drop the file into your public folder and deploy.
5. Fetch it yourself to verify. Open the URL in a browser or run curl. If you get a 404 or an HTML error page, it's not live.
6. Check that AI can actually reach the rest of your site too. An llms.txt pointing at pages that GPTBot is blocked from crawling is wasted effort. Scan your site free and SuperSEO.sh will check your robots.txt per AI bot, your JS-rendering risk and your structured data in one pass.

That's the whole job. No build tooling, no schema validation, no registration with anyone.

Does llms.txt actually get used?

The honest answer: llms.txt is a young, unofficial standard — proposed by Jeremy Howard in 2024 — but it has been picked up fast by AI-focused companies, and the cost of adding one is close to zero.

What we know: major AI companies including Anthropic publish llms.txt files for their own documentation, and LLM-powered tools that fetch websites increasingly request /llms.txt as a first stop because it's the cheapest way to understand a site. What we don't know: exactly how much weight each assistant gives it, or whether ChatGPT will cite you more often because of one file. No vendor can promise that, and you should distrust anyone who does.

What we can say confidently is the risk/reward math. Writing an llms.txt takes ten minutes and costs nothing. If it helps even occasionally — an agent reading your docs correctly, a model describing your product with your own words instead of a hallucinated guess — it has already paid for itself. It's one of several signals that make your site easier for AI to understand, alongside the others that determine whether you get cited by ChatGPT at all.

FAQ

Where does the llms.txt file go?

The llms.txt file goes at the root of your domain, so it resolves to https://yoursite.com/llms.txt. Subdirectories and subdomains can have their own, but the root-level file is the one tools look for first.

What format rules does llms.txt follow?

llms.txt is standard markdown with a fixed skeleton: one H1 (your name), one blockquote (your summary), then H2 sections containing link lists. There is no central validator, but sticking to that structure is what parsers expect.

How often should I update my llms.txt?

Update your llms.txt whenever your important pages change — a new pricing page, a rewritten docs section, a flagship blog post. For most indie sites that's a few times a year, not a weekly chore.

Does Google care about llms.txt?

Google has said it does not use llms.txt for search ranking, and there is no evidence it affects your position in search results. llms.txt is aimed at AI assistants, not at Googlebot — treat it as an AI visibility tool, not an SEO ranking factor.

Does llms.txt replace SEO?

No — llms.txt is an addition to SEO, not a replacement for it. Google still sends the overwhelming majority of search traffic, and llms.txt does nothing there. It covers a different channel: being understood and cited when people ask AI assistants instead of typing into a search box. You need both.

AI visibility is layered: let the crawlers in, make your content extractable, then hand the models a map. The map takes ten minutes — and scan your site free to find out whether the other layers are already costing you mentions.