llms.txt is the map of your site written for assistants rather than people. Nobody ever opens
that URL, so it is written once and then quietly rots: pages get renamed, sections disappear,
and the only reader who notices is a machine that cannot tell you.
Public files only. The address you type is not stored.
What a working llms.txt looks like
# Isla Verde Property
> Independent buyer guides for Isla Verde: what condos cost by district, the fees at completion, the title checks that matter.
## Buyer guides
- [Fees and taxes at completion](https://example.com/guides/fees-and-taxes-at-completion/): every fee a buyer pays, with worked examples at three prices
- [Service charges in condo buildings](https://example.com/guides/service-charges/): what the monthly charge covers and how it is set
One H1 naming the site. One blockquote saying what it is. Sections of links, each with a line
explaining what the reader will find. Nothing else is required, and nothing else is read.
The three ways it goes wrong
It does not exist. The most common case by far. An assistant that wants a map finds nothing and reads whatever your HTML happens to give it.
It was written by hand. A hand-written map is accurate on the day it is written. Generate it from the same content that builds your pages and it cannot drift.
It points at pages that moved. Links rot silently here because no human opens the file. This is what the check above is for.
Frequently asked questions
What is llms.txt?
A plain text file at the root of a site, /llms.txt, that lists the pages worth reading with one line each. It is a map written for programs: an assistant reads it instead of crawling the whole site, and decides from it what you have. The format asks for one heading naming the site, a short summary, and sections of links.
Does llms.txt do anything for rankings?
Not for Google rankings. It matters to assistants that fetch a site to answer a question: with a map they read the right pages, without one they read whatever the HTML gives them and often quote the wrong thing. No engine has promised to honour it, and anyone who tells you otherwise is selling something.
Why check it at all?
Because nobody ever opens that URL. The file is written once and the site keeps moving: pages get renamed, sections disappear. Six months later half the map leads nowhere and the only reader affected is a machine that cannot complain.
What does this check look at?
Whether the file exists, whether it starts with a heading as the format asks, whether it carries a summary and sections, how many links it lists, and where up to twelve of those links actually lead: a page, a redirect, or nothing. It also looks for llms-full.txt beside it.
Do you store the address I check?
No. The check runs while the page loads and nothing is written down. There is no account and no cookie for it.
How do I write one?
Generate it from your own content rather than by hand, so it cannot drift. The free OperStack Astro starter builds llms.txt and llms-full.txt from the corpus on every build, which is why its map cannot go stale. Any generator that reads your sitemap will do the same job.