Emerging
llms.txt and agent.json: the new front door
For thirty years the first file a crawler asked for was robots.txt. A second file is now appearing beside it — a plain summary of the site, written for language models rather than search indexes. Almost nobody in Tasmania has one yet.
Two files, two jobs
llms.txt sits at the root of your domain, like robots.txt, and is a short plain-text summary: who you are, what you do, where you work, and where the important pages are. The idea, proposed in 2024, is that a model that finds it spends less time guessing.
agent.json goes under /.well-known/ and describes what an automated assistant may do with your site — what data you publish, in what format, and how to ask for it. It is early and informal, but it is how a site starts being usable by an agent rather than merely readable by a person.
How common are they really?
Not common. Of the 1,045 Tasmanian business sites we checked, 44 have a genuine llms.txt — 4.2%. Several dozen more return a page at that address, but it is the site's 404 page pretending to be a file, which is worse than not having one: it teaches nothing and wastes a request.
Being early is the only interesting thing about it. It is not a ranking factor. No assistant has promised to read it. But it costs ten minutes and it is the cheapest way to be unambiguous about what your business is.
What goes in llms.txt
# Example Plumbing — St Helens, Tasmania
> Plumber serving St Helens, Binalong Bay and Scamander.
> Emergency callouts, hot water, blocked drains. Phone 03 6257 1300.
## Key pages
- [Services](https://example.com.au/services): what we fix and what we do not
- [Service area](https://example.com.au/areas): towns we travel to
- [Contact](https://example.com.au/contact): phone, hours, after-hours number
## Facts
- ABN 00 000 000 000
- Licences: Tasmanian plumbing licence XXXXX
- Hours: Mon-Fri 07:00-16:00, emergency after hours
Two rules matter: every claim must be true and visible elsewhere on the site, and the file must be plain text at exactly /llms.txt. Not llms.txt.html, not a page that redirects.
Is it worth doing now?
Honestly: it is worth ten minutes, not a project. The businesses that will benefit are those whose customers have already started asking assistants instead of scrolling. If that is not yet your trade, the five fields in the previous article matter more.
What it does buy you is a clean, factual statement of the business in one place — which is also the best brief you will ever write for a designer, a directory listing or a journalist.
What we do here
This site runs its own llms.txt and an agent-facing summary of what it publishes. We build the same for clients as part of the audit — and if you only want the two files, that is a short job, quoted as one.
Questions people ask
Does llms.txt improve my Google ranking?
No. It is not a search ranking signal. It is a summary for language models that choose to fetch it.
Will ChatGPT read my llms.txt?
Only if it fetches the file, which depends on the tool and the question. There is no guarantee. It removes ambiguity for the tools that do look.
Can I just copy another business llms.txt?
The structure, yes. The content has to be yours and true — a copied file with someone else details is worse than none.