llms.txt is a plain text file you place at the root of your website (yourdomain.com/llms.txt) that gives AI systems a clean, structured map of your most important content — similar to how robots.txt tells search engines where to crawl. Instead of making ChatGPT, Claude, or Perplexity sift through your cluttered HTML, navigation menus, and ad scripts, llms.txt hands them a curated list of the pages that actually matter. In 2026, that matters because a growing share of your traffic now comes from AI answers instead of blue links — and the sites that make life easy for AI get cited more often.
This guide covers what llms.txt is, where it came from, what to put in it, and exactly how to create and upload one in under ten minutes. llms.txt is one of the core signals covered in our generative engine optimization guide, alongside schema, citations, and content structure.
What Is llms.txt?
llms.txt is a markdown file that lives at the root of your domain — so if your site is acme.com, the file sits at acme.com/llms.txt. It is written in plain text with a simple structure: a site name, a short description, and a list of links to your most important pages grouped into sections.
The goal is simple. AI models have limited context windows (the amount of text they can read at once). When an AI system visits your website to answer a question, it cannot — and should not — read every page. llms.txt gives it a pre-selected, human-curated list of the content that best represents your business, so the model can quickly find what it needs and cite you accurately.
Think of it as a restaurant menu for AI. Your full website is the kitchen, with everything you have ever cooked. llms.txt is the menu — the curated selection you want visitors to actually order from.
Where Did llms.txt Come From?
The llms.txt standard was proposed in September 2024 by Jeremy Howard, co-founder of fast.ai and Answer.AI, and one of the most respected figures in applied AI research. Howard noticed that AI systems were increasingly being used to read websites on behalf of users, but most websites were not designed for machine consumption.
The full specification lives at llmstxt.org, which is the official home of the standard. The proposal borrows the naming convention from robots.txt — the file websites have used since 1994 to tell search engine crawlers which pages they can access — but solves a different problem. robots.txt controls access; llms.txt controls comprehension.
According to llmstxt.org, the file is designed to "provide information to help LLMs use a website at inference time" — meaning at the exact moment an AI is trying to answer a user's question about your content, not months earlier during training.
Since the proposal, adoption has grown quickly. According to Writesonic's llms.txt adoption tracker, thousands of sites including Anthropic, Stripe, Zapier, Cloudflare, and Perplexity have published llms.txt files — well before any AI platform officially required them.
llms.txt vs robots.txt vs sitemap.xml
These three files are often confused. They all sit at the root of your domain and they all control how machines interact with your site. But they do very different jobs.
| llms.txt | robots.txt | sitemap.xml | |
|---|---|---|---|
| Purpose | Curate content for AI to read and cite | Block or allow crawlers from accessing pages | List every URL for search engines to index |
| Audience | LLMs (ChatGPT, Claude, Perplexity, Gemini) | Search engine bots, scrapers | Search engine bots |
| Required? | No — proposed standard, voluntary | No — but universally supported since 1994 | No — but standard SEO practice |
| Format | Markdown with links and short descriptions | Plain text rules (User-agent, Disallow) | XML with <url> entries |
| Location | /llms.txt | /robots.txt | /sitemap.xml |
| Typical length | 20–200 curated links | 5–30 lines of rules | Every URL on the site (can be thousands) |
| Goal | Make content easy to understand | Control what gets crawled | Help search engines discover everything |
The simplest way to remember it: robots.txt is a bouncer, sitemap.xml is a phonebook, and llms.txt is a menu.
Why llms.txt Matters for AI Visibility in 2026
The shift from search engines to AI answer engines is no longer a prediction — it is the default for a growing share of users. According to Semrush, ChatGPT referral traffic to websites grew by more than 400% between 2024 and 2026, and a meaningful portion of queries that used to go to Google now end inside an AI chat window.
When AI systems cannot parse your site cleanly, three things happen:
- You get misquoted. The model grabs outdated or irrelevant content and puts the wrong words in your mouth.
- You get skipped. If another site's content is easier to read, the AI cites them instead.
- You lose the traffic. AI answers with citations drive clicks; answers without them do not.
According to SearchEngineLand, well-structured content that is easy for LLMs to parse is significantly more likely to appear in AI citations — and llms.txt is one of the fastest, cheapest signals you can add.
Unlike SEO tactics that take months to show results, llms.txt is a one-file change. You can publish one in an afternoon. For more on the broader strategy, see GEO vs SEO: what's actually different and how to get cited by ChatGPT.
What Goes in an llms.txt File
The format is deliberately minimal. A valid llms.txt file has:
- An H1 heading with your site or business name — one line, no fancy formatting.
- A blockquote description — a 1–3 sentence summary of what your business does and who it serves.
- Optional context paragraphs — short explanations of anything an AI should know before reading your links.
- H2 sections with grouped links — each link is a markdown list item with the URL and a short description.
- Optional
## Optionalsection — secondary links you want the AI to know about but that are not critical.
Here is what a complete, real-world example looks like:
# ACME Coffee Roasters
> ACME Coffee Roasters is a speciality coffee company based in Portland, Oregon. We source single-origin beans directly from farms in Ethiopia, Colombia, and Guatemala and ship freshly roasted coffee to customers in the United States and Canada.
We have been roasting since 2012. Our team includes three Q-graders and a head roaster with 15 years of experience. We publish tasting notes, brewing guides, and farm profiles on our blog.
## Core pages
- [Homepage](https://acme-coffee.com): Main landing page with current featured beans
- [About us](https://acme-coffee.com/about): Our story, team, and sourcing philosophy
- [Shop all coffee](https://acme-coffee.com/shop): Complete catalogue of single-origin and blend coffees
- [Subscriptions](https://acme-coffee.com/subscribe): Monthly and bi-weekly delivery options
- [Wholesale](https://acme-coffee.com/wholesale): Coffee for cafes, restaurants, and offices
## Brewing guides
- [Pour over guide](https://acme-coffee.com/guides/pour-over): Step-by-step V60 method
- [French press guide](https://acme-coffee.com/guides/french-press): Ratios, grind size, and timing
- [Espresso at home](https://acme-coffee.com/guides/espresso): Equipment, dosing, and dialling in
