Websites have had files for decades that aren't meant for visitors but for machines. robots.txt tells crawlers what they may read, sitemap.xml tells search engines which pages exist. llms.txt is the newest in that line, and the target audience is new: the language models behind ChatGPT, Claude and Perplexity.

01 / the fileWhat's in llms.txt

An llms.txt is a short, readable text file at the fixed address yourdomain.com/llms.txt. The proposal comes from llmstxt.org and is deliberately kept simple: plain text with some structure, no code. The standard parts:

• one sentence stating what the site is and who it's for;
• the core pages with full links: the offering, the prices, the explanatory or knowledge pages and the contact details;
• half a sentence per link about what's on it;
• where applicable: documentation or frequently asked questions.

The difference with robots.txt lies in the direction. robots.txt says what a machine may not do; llms.txt says what a machine should particularly see. It's a guide, not a lock on the door.

02 / the nuanceWhat Google thinks of it, and the rest

Honesty about the file's status is in order here. Google explicitly states that no special markup or AI files are required for AI Overviews: Google's AI features rest on the regular search systems, and good content for people is the whole story there.

Outside Google it's different. Systems like ChatGPT, Claude and Perplexity do read such files when they're present, and a clear guide saves such a system some searching: it finds what you offer and where it is more quickly. Nowhere is it proven decisive. See it as the flag on the building, not the foundation underneath. The foundation remains content that genuinely answers a question, and how you write that is covered in how to write content AI can cite.

llms.txt is a guide for machines: a small effort, no miracle cure, and certainly no substitute for good content.

03 / in practiceHow to make one

Making one takes less time than reading this article:

1. Open a text editor and write your company name at the top, with one sentence about what you do and for whom.
2. Below that, list your core pages: per line the page name, the full link and half a sentence about what's on it.
3. Save the file as llms.txt.
4. Have it placed in your site's root directory, next to robots.txt, so it's accessible at yourdomain.com/llms.txt.

Two points of attention. Keep it up to date: a guide pointing to pages that no longer exist works against you. And don't put anything in it that isn't also simply on the site; the file is public and should be a summary, not a shadow version of your site.

We ourselves also run an llms.txt on datadriftdigital.nl, with exactly that structure: what the company does, and where the core pages are. Not because the file works wonders, but because it was ten minutes of work, and sites that write about machine readability may be expected to apply it themselves.

04 / the limitWhen you can safely skip it

If your site isn't there content-wise yet, llms.txt isn't the place to start. A guide pointing to pages that don't answer your customers' questions only steers a machine to the wrong thing faster. Start with the pages themselves instead, and with the question of whether AI systems are allowed to read your site. If you then also want to help machines understand what's on those pages, that's the territory of schema markup; more on that in does schema markup help with AI visibility.

If your site is small and clear, with five pages and a straightforward structure, llms.txt adds little: there's nothing to summarise that a machine can't already see for itself in one go. The file pays off most on sites with many pages, where the core would otherwise get lost among the rest.

This text was produced with AI assistance and checked and approved by a human before publication.