If you've spent any time looking at GEO checklists lately, you've probably run into llms.txt and wondered whether it's a real thing or just another file someone invented to sell you a course. Fair question. So let's be straight about it: llms.txt is a young, voluntary convention, no search engine requires it, and publishing one will not magically get you cited tomorrow.
It's still worth doing. Here's why, and how.
The one-paragraph version
llms.txt is a Markdown file you put at the root of your domain (yourdomain.com/llms.txt) that lists the pages on your site you actually want an AI model to read, with a sentence of context for each. Think of it as a curated table of contents written for machines instead of a sitemap that dumps every URL you've ever published. There's an optional companion, llms-full.txt, that inlines the full text of those pages so a model doesn't even have to fetch them separately.
That's the whole idea. A human-readable, AI-readable summary of what matters on your site.
Why bother, honestly
Large language models don't read your site the way Google's crawler does. When someone asks ChatGPT or Perplexity a question, the model has a tight budget for how much it'll fetch and process before answering. If it lands on your homepage and hits a wall of navigation, cookie banners, and three carousels before any actual content, that's a bad first impression you don't get to redo.
llms.txt skips the mess. It says: here are the five pages that answer the questions people ask about us, in order, each with a one-line description. You're doing the model's filtering for it, which makes your content cheaper to use, and cheaper to use tends to mean more likely to get used.
Is the payoff guaranteed? No. Adoption is still early and the engines treat it as a hint, not a directive. But the file takes twenty minutes to write, costs nothing, and signals that you're thinking about how machines read you. We've seen enough sites add it as part of a broader GEO cleanup that we flag its absence in every audit.
What a good one looks like
The format is just Markdown. An H1 with your site name, an optional blockquote summary, then ## sections of links. Each link gets a short description after a colon.
# Acme Analytics
> Acme Analytics is a privacy-first product analytics tool for SaaS teams.
## Core pages
- [What Acme does](https://acme.com/product): Overview of features and who it's for.
- [Pricing](https://acme.com/pricing): Plans, limits, and what's included free.
- [Docs: getting started](https://acme.com/docs/start): Install and send your first event.
## Guides
- [GDPR-safe analytics](https://acme.com/blog/gdpr): How we handle EU data without cookies.
Keep it tight. Five to fifteen links is plenty for most sites. The goal is to point at your best, most answer-shaped content, not to mirror your whole sitemap. If a page wouldn't help someone who asked a real question about your business, leave it out.
How to publish it
- Write the file in any text editor and save it as
llms.txt. - Drop it at your web root so it resolves at
https://yourdomain.com/llms.txt. On most static hosts that means yourpublic/or root folder; on a CMS you may need a redirect or a small route. - Confirm it loads in a browser and returns
text/plain(or at least plain text, not HTML). - Optional: add
llms-full.txtwith the full Markdown of those pages inlined, if your content is stable enough to maintain a second copy.
That last point is the catch with llms-full.txt: it's a duplicate that can drift out of date. If you can't commit to keeping it fresh, skip it. A correct llms.txt beats a stale llms-full.txt.
Common mistakes
- Listing everything. A 200-link file isn't curation, it's a sitemap with extra steps. Pick your strongest pages.
- Dead or redirecting links. Models follow them and waste their budget. Check them the same way you'd check a sitemap.
- Serving it as HTML. It's meant to be plain text. A styled page defeats the purpose.
- Writing it once and forgetting it. When you publish a great new explainer, add it. When you kill a page, remove it.
Where it fits in the bigger picture
llms.txt is one signal among several. It helps an engine find your good content, but the content still has to be good: clearly structured, dated, and quotable. So treat this as a companion to the rest of your GEO work, not a substitute for it. If you haven't already, make sure you're letting AI crawlers reach your pages in the first place and that your key pages carry structured data. Those two do more heavy lifting than llms.txt ever will.
FAQ
Is llms.txt an official standard?
No. It's a community proposal that some sites and tools have adopted. Treat it as a low-cost best practice, not a requirement.
Will llms.txt get my site into ChatGPT's answers?
Not on its own. It makes your best content easier to find and cheaper to process, which helps, but answerability, structured data, and being crawlable matter more.
Do I need llms-full.txt too?
Only if you can keep it current. A stale full-text copy is worse than none. Start with llms.txt and add the full version later if it's worth maintaining.
Where exactly does the file go?
At your domain root: https://yourdomain.com/llms.txt. Not in a subfolder.
Key Takeaways
llms.txtis a Markdown file at your domain root (yourdomain.com/llms.txt) that lists the pages you want AI models to read, each with a one-line description, acting as a curated table of contents rather than a full sitemap.- It is a young, voluntary convention that no search engine requires; engines treat it as a hint, so it will not get a site cited on its own.
- Keep the file tight at roughly five to fifteen links, serve it as plain text rather than HTML, and remove dead or redirecting links so models don't waste their fetch budget.
- The optional
llms-full.txtinlines the full text of those pages, but a stale full-text copy is worse than none, so only add it if you can keep it current. - Crawlability and structured data do more for AI visibility than
llms.txt, so treat the file as a companion to the rest of a GEO cleanup, not a substitute.
Want to know whether your site already publishes a valid llms.txt, and how the rest of your GEO signals stack up? Run a free audit and we'll check it in seconds. You can also browse the other GEO explainers for the checks that matter more.
