Some patients now ask an AI assistant which doctor to see. Others ask one to find a clinic’s opening hours, or to send it a question. To do either, the assistant reads websites much as a search engine does, except that it is looking for facts it can repeat. A file called llms.txt is one small way to hand it those facts directly.
What the file is
llms.txt is a plain text file at the root of a website, at an address such as yourpractice.com/llms.txt. It is written in Markdown, the simple format used for notes and documentation, and it follows a public proposal published at llmstxt.org.
It is a proposal, not an official standard, and each assistant decides for itself whether to read it. It costs nothing to have, and it gives the ones that do read it a clear summary instead of a guess.
What goes in it, for a practice
For a medical practice, a useful llms.txt is short. It says:
- who you are: your name, your specialty and your qualifications;
- what you treat, in plain words;
- where you practise, with your address and opening hours;
- which languages you see patients in;
- how to get in touch or book;
- and where on the website each of these is explained in full.
It should say nothing the website does not also say. It is a summary of the site, not a second one.
The other things that help
The file on its own is a small part of the picture. Three other things matter at least as much.
- Structured data. Labels in the page’s code, written in a shared vocabulary called schema.org, that say “this is a medical practice”, with its specialty, address, hours and languages, so a machine has nothing to guess.
- Not turning crawlers away. A website’s robots.txt file can block the programs that AI companies use to read the web. Some sites block them without anyone meaning to.
- Pages that work without scripts, and labelled buttons. When an assistant visits on a patient’s behalf, to find a number or send a question, it needs pages that read fully without running code, and buttons and forms whose purpose is labelled. The same habits make a site usable with a screen reader.
How to check your own website
Type your website’s address into a browser followed by /llms.txt. If you see a page of plain text about your practice, you have one; if you see an error page, you do not.
You can also open your address followed by /robots.txt. A line that begins “Disallow: /” under the name of a crawler means that crawler is being turned away. Anything beyond that is a question for whoever looks after your website.
What no one can promise
No one can promise what an assistant will say about you, or whether it will suggest you. What a website can do is make sure that when an assistant looks, it finds accurate facts, stated plainly, in a form it can read. That part is within your control.
Every website we look after serves an llms.txt, carries structured data describing the practice, and turns no AI crawler away. This one does too: you can read ours at nodsites.com/llms.txt.