In short: llms.txt is a text file at the root of a website giving AI models a curated overview of your content. No major model provider has committed to using it, and Google has said it does not. Still, we published our own at synlig.ai/llms.txt. Our conclusion is sober: Neither hype nor necessity, but cheap insurance with zero downside, which also happens to enforce good content discipline.
What is llms.txt?
The proposal came from developer Jeremy Howard in 2024: A Markdown file at yourdomain.com/llms.txt telling language models who you are and which pages matter most, with a short description per link. The idea is to do for AI models what robots.txt does for crawlers and sitemap.xml does for search engines: Give the machines a clean map instead of letting them guess their way through menus, popups and decoration.
Who actually uses it?
Here is where hype and reality part ways. A number of AI tools and documentation platforms have adopted the format, and many websites have published the file. But no major model provider has confirmed fetching llms.txt systematically in production, and Google has explicitly stated the file is not used in its systems. Today's status: A sensible proposed standard waiting for the big players to commit.
Which means llms.txt does not measurably move your AI visibility today. What moves it is still the foundation: Technical accessibility for AI crawlers, structured data and content worth citing.
Why did we publish one anyway?
- The cost benefit ratio is extremely skewed. The file took under an hour to set up, interferes with nothing, and sits ready the day a model provider starts fetching it. Insurance with a zero premium is worth accepting.
- The discipline it forces has value in itself. Writing an llms.txt means answering «which ten pages best explain what we offer?». That exercise reveals gaps in your content structure that also affect regular SEO and GEO.
- We make a living from being early. We advise our clients to position themselves in AI surfaces before their competitors do. We should take the same low stakes bets ourselves.
Our llms.txt is generated automatically by the website's build system, so it stays current as new content is published. See it at synlig.ai/llms.txt.
Should you create one?
Yes, if it is practically free for you, meaning someone can set it up in under an hour. No, if it steals time from the foundation. A business that does not yet have structured data, clear answer-first pages or technical access for AI crawlers should spend the hours there first. llms.txt is the cherry, not the cake.
Want to know what actually moves your visibility?
We measure how businesses get recommended in ChatGPT, Gemini, Google AI and the other models, and which sources drive the answers. Book a demo and we will show you where you stand and what to prioritize, with llms.txt in its right place in the queue.