llms.txt Explained: What It Is, Whether It Works, and How to Write One
llms.txt is a plain-text file at the root of your site that summarises what the site is and what is on it, in Markdown, for AI crawlers and answer engines. It is an emerging convention rather than a standard any engine has committed to honouring — so treat it as cheap insurance, not a ranking lever.
The idea borrows from robots.txt. Where robots.txt says what a crawler may access, llms.txt says what the site actually contains — a short description of the business, then structured lists of the important pages with one line each. The argument for it is that a model reading a clean summary is less likely to misunderstand a site than one inferring from navigation and marketing copy.
- It is a convention, not a standard. No major engine has committed to reading it.
- It costs almost nothing to publish, which is the whole argument for doing it.
- Generate it from your own content so it cannot drift out of date.
- An empty or stale field is worse than omitting it — it reads as a gap in the business.
What goes in it
- An H1 with your business name, then a one-line blockquote summary.
- A short paragraph saying plainly what the business does and where it operates.
- Sections as H2s — services, products, pricing, FAQs, articles — each a list of links with a one-line description.
- Contact details, so the model has a route to attribute.
Keep descriptions factual. This is not a place for positioning language; it is a place to be unambiguous about what exists.
The honest state of adoption
Worth being direct, because there is a lot of confident writing about this. No major AI provider has publicly committed to reading llms.txt as a ranking or retrieval input. Publishing one is not going to move you into an answer on its own. What it does is guarantee that if a crawler does read it, the summary it finds is accurate, current and written by you rather than inferred from your homepage.
That is a reasonable trade for an afternoon of work. It is not a reasonable basis for a retainer sold on llms.txt alone.
Generate it, do not write it
A hand-written llms.txt is accurate on the day it is published and wrong within a quarter. Generating it from the same data that produces your site means it updates itself when you add a service, publish an article or change a price. On this site, /llms.txt is produced from the same content providers that render the pages — services, work, FAQs, packages, products and the journal.
Frequently asked questions
Where does llms.txt go?
At the root of the domain, served as plain text — the same place as robots.txt. Some sites also publish a fuller llms-full.txt containing the complete content rather than a summary.
Does Google or OpenAI actually read it?
Neither has publicly committed to it. Anyone telling you otherwise is reporting a hope rather than a documented behaviour. Publish it because it is cheap and accurate, not because it is confirmed to work.
Is it the same as robots.txt?
No. robots.txt controls access — what a crawler may fetch. llms.txt describes content — what the site contains. They do different jobs and you want both: explicit crawler permissions in robots.txt, and an accurate summary in llms.txt.
Should we block AI crawlers instead?
That is a genuine strategic choice, not an oversight. A publisher whose content is the product may rationally block them. A business that wants to be found and recommended should not — being unreadable to the systems answering your buyers’ questions is an odd position to adopt deliberately.
You can read this site’s llms.txt, which is generated rather than written. Checking whether yours is being read correctly is part of the AI search visibility audit.
