In short: llms.txt is a Markdown file at the root of a website that gives AI systems a compact, curated list of its most important pages and content, similar to how robots.txt works for crawlers and sitemap.xml works for search engines. ChatGPT, Gemini, and Google haven't officially confirmed they use it yet, though some developer tools already read it actively. Okapi Digital breaks down what llms.txt actually delivers today and how to set one up correctly.
Anyone reading about ChatGPT SEO or AEO keeps running into the same tip: "add an llms.txt file." Search what is llms.txt and most answers skip the honest part: what the file can and can't do. Consider this llms.txt explained without the hype: the format, a real example, and a plain answer on which AI systems actually read it today.
What Is llms.txt, Exactly?
llms.txt is a plain text, or more precisely Markdown, file that lives at /llms.txt at the root of a domain. AI researcher Jeremy Howard (Answer.AI) proposed the format in September 2024, with a simple rationale: language models only hold a limited context window, and a normal HTML page is packed with navigation, ad slots, and layout code that's worthless to a model. llms.txt instead offers a short, clean reading list of a site's key content, in plain text.
The name deliberately echoes robots.txt, but it follows a different logic. robots.txt grants or blocks crawling of specific paths. llms.txt instead actively recommends which pages matter most to an AI system, framed as a curated list rather than a rulebook.
How Is an llms.txt File Structured?
The format follows a fixed, simple Markdown structure:
# Okapi Digital
> Frankfurt-based digital agency for web design, SEO, branding, and AI automation.
## Services
- [SEO Agency](https://okapidigital.io/services/seo): Local and technical SEO, AEO, and GEO
- [Web Design & Development](https://okapidigital.io/services/website-development): Websites built on Webflow, Framer, and Next.js
## Optional
- [Blog](https://okapidigital.io/news): Guides on SEO, AI automation, and web design
The structure has four fixed parts: an H1 heading with the business or project name, a short blockquote as a one-sentence summary, several H2 sections grouping links by topic with a short description each, and an optional "Optional" section for less central content. Every link gets a brief, concrete sentence about what the destination page actually covers, not just the page title.
Do ChatGPT, Gemini, and Google Already Use the llms.txt File?
As of 2026, none of the major AI companies have officially confirmed that their crawlers or search features systematically read llms.txt or use it for rankings or citations.
| System | Official llms.txt support |
|---|---|
| ChatGPT search (OAI-SearchBot, GPTBot) | Not officially confirmed |
| Google (Search, AI Overview) | Not officially confirmed |
| Gemini / Google-Extended | Not officially confirmed |
| Various AI coding assistants and documentation tools | Several already read the file actively when given a URL |
What that means in practice: llms.txt is not a confirmed ranking factor and no guarantee of a citation in ChatGPT or Google AI Overview. At the same time, adding the file costs almost no effort, and it does no harm. That's why, in AEO practice, it counts as a sensible, low-risk addition rather than a silver bullet. For the ranking factors that are actually confirmed to matter for ChatGPT, Bing indexing, crawler access, and answer structure, see our guide to ranking in ChatGPT.
llms.txt vs. robots.txt vs. sitemap.xml: What's the Difference?
The robots.txt vs llms.txt framing is misleading, since the two aren't competing for the same job. All three files sit at the root of a website, but each does a different job:
| File | Audience | Job |
|---|---|---|
robots.txt | All crawlers (Google, Bing, OAI-SearchBot, GPTBot) | Allows or blocks crawling of specific paths |
sitemap.xml | Search engines | Lists every indexable page in full |
llms.txt | AI language models with limited context | Curates a short, prioritized reading list of key content |
None of the three files makes another one unnecessary. A website with a correct robots.txt, an up-to-date sitemap.xml, and a clean llms.txt covers three distinct technical layers: access, completeness, and prioritization.
How to Create Your Own llms.txt File: 4 Steps
The steps below show how to create llms.txt without needing a developer on hand.
Step 1: Pick Your Most Important Pages
Choose five to fifteen pages that describe your offering, pricing, and key content most clearly: homepage, service pages, core blog posts, contact page. More isn't automatically better here.
Step 2: Write the File in Markdown
Follow the structure in the example above: an H1, a short blockquote, and H2 sections with links plus a brief description each. Write every description so it makes sense even without clicking the link.
Step 3: Publish the File at /llms.txt
Upload the file to your server's root directory, reachable at yourdomain.com/llms.txt. On a Next.js site, set up a simple route that serves the Markdown content as plain text. On WordPress, a small plugin or a static file in the root directory handles it.
Step 4: Check It's Reachable and Keep It Current
Load the URL directly in a browser and confirm the content is readable without a login. Update the file whenever key pages, pricing, or services change, the same way you'd keep a sitemap current.
What Are the Most Common llms.txt Mistakes?
In practice, the same handful of mistakes keep an otherwise carefully built llms.txt from doing anything useful:
| Mistake | Why it kills the effect |
|---|---|
| File isn't at the domain root (e.g., under /blog/llms.txt) | AI systems only check /llms.txt, so a file anywhere else simply never gets found |
| File sits behind a login or paywall | Crawlers can't sign in, so the content stays invisible to them |
| Every single page gets listed, like a sitemap | The point of the file is prioritization; a full list adds nothing sitemap.xml doesn't already cover |
| The blockquote summary is too vague | An AI system can't pull a concrete fact about the business out of "we're innovative" |
| The file gets written once and never touched again | Stale links and old prices damage trust the moment anyone actually fetches it |
| llms.txt is treated as the only AEO move | Without an open robots.txt, structured data, and answer-first content, its effect stays small |
That last point matters most: llms.txt only does real work alongside the rest of the technical groundwork, not as a standalone fix.
Who Should Set Up llms.txt Right Now?
llms.txt makes the most sense for businesses already working on other GEO/AEO basics, such as Bing indexing, structured data, and answer-first content, as described in our guide to AEO in ChatGPT, Gemini & more. For those businesses, llms.txt is a small, quick add-on to a broader package, not the only move. If your robots.txt or sitemap still isn't clean, fix those fundamentals first; llms.txt can't make much difference until they are.
Conclusion
llms.txt is a simple, free Markdown file that gives AI systems a curated overview of your most important content. The major AI providers haven't officially confirmed they use it yet, so it's not a guaranteed ranking factor, but it is a low-risk addition alongside robots.txt, a sitemap, and schema markup. Okapi Digital's SEO Ultra plan builds AEO and GEO moves like this one in by default. See our SEO services page for details.
Want to know how your website is technically set up for ChatGPT, Gemini, and Google AI Overview today? Book a free discovery call. We'll check your robots.txt, sitemap, schema, and llms.txt, and show you exactly where the gaps are.



