What Is llms.txt, and Does Your Cayman Business Need One?

If you have been anywhere near a conversation about AI and websites in the last year, someone has probably told you that you need an llms.txt file. They may not have been able to tell you what it does.

Here is the honest version, including the part most people selling it will skip: what it is, what it demonstrably does today, what it does not do, and why we still put one on every site we build.

What the file actually is

An llms.txt is a single plain-text file that sits at the root of your website — yourbusiness.ky/llms.txt — and describes your business in language written for a machine to read.

It was proposed in late 2024 as a convention, in the same spirit as robots.txt. The reasoning behind it is straightforward. A modern website is a difficult thing for a model to parse: the useful facts are buried under navigation, cookie banners, tracking scripts, image carousels and layout markup. A large language model working with a limited context window has to wade through all of it to find out what time you close.

An llms.txt skips the wading. It is your business, stated plainly, in Markdown, with no markup to strip: who you are, what you sell, where you are, what makes you different, and links to the pages worth reading in full.

The part nobody selling this will tell you

It is a proposed standard. As of this writing, no major AI company has publicly committed to reading it. Not OpenAI, not Google, not Anthropic. There is no announcement to point at, and anybody telling you that ChatGPT reads your llms.txt is telling you something they cannot demonstrate.

We think you should know that before deciding, and we would rather say it ourselves than have you find out later and wonder what else we oversold.

So why do we ship one on every build?

Because the cost is close to zero and the asymmetry is absurd. Writing one takes an afternoon. It is a text file — it cannot slow your site down, it cannot break your design, it cannot conflict with anything. If the convention is adopted, you were early at no cost. If it is never adopted, you spent an afternoon.

That is not a thrilling argument. It is an honest one, and it is the same reasoning that had people adding structured data in 2012, years before Google did much with it.

What it is not

Three misconceptions worth clearing up, because each one leads somewhere expensive:

• It is not a ranking factor. Nobody ranks higher on Google because of this file. If someone quotes you a monthly retainer built around llms.txt, that is a retainer built on a text file.
• It is not a replacement for structured data. Schema markup is read by Google today, right now, with documented effects on how your listing appears. llms.txt is a bet on tomorrow. Do the first one first.
• It is not a way to hide things or game anything. It is a public file at a public URL. Anything you put in it, your competitor can read.

What goes in a good one

The format is loose by design: a Markdown document with an H1 for your business name, a short summary, and sections of annotated links. What separates a useful one from a decorative one is the writing.

Most of the ones we see in the wild are a table of contents — a bare list of page URLs with their titles. That is nearly worthless, because a crawler could already get that from your sitemap.

A useful one states facts a model cannot infer from your page titles:

• Where you actually are, in words a person would use. “George Town, Grand Cayman, Cayman Islands” — not just “Cayman”.
• What you sell, in the words customers use rather than your internal names for things.
• Who you serve and who you do not. Being explicit that you work with small local businesses is more useful than claiming you serve everyone.
• The things that are true and checkable: how long you have been trading, memberships, the industries you have actually worked in.
• Your other official profiles, so a model reading about you elsewhere can connect them to the same business.

What ours looks like

You can read the whole thing at toctoc.ky/llms.txt — we would rather show you than describe it. A few decisions in there are deliberate and worth stealing:

It is generated from the site, not maintained as a separate document. A file that has to be updated by hand is a file that is wrong within two months, and a confidently wrong description of your business is worse than none at all.

It names the limits of our own claims. Where we describe AI assistants naming our clients, the file says those were recorded sessions on specific dates, that assistants are non-deterministic, and that they are not guaranteed positions. If a model does read it, we would rather it repeat an accurate sentence about us than an inflated one.

It declares our official profiles, including the note that our Instagram handle changed in August 2026 and the old account is a redirect. That kind of small, boring fact is exactly what stops a model from treating one business as two.

Where it sits in the order of things

If you are going to spend a limited budget on being readable by AI, spend it in this order:

1. Get your content out of PDFs and images. Real, selectable text on real pages. Nothing below this matters until this is true.
2. Add structured data. Your business type, address, coordinates, hours, services, prices. Read by Google today.
3. Make your details consistent everywhere. Website, Google Business Profile, Apple Maps, social. One version of the truth.
4. Then add llms.txt. Cheap, fast, and possibly early.

A business that does the first three and skips the fourth is in good shape. A business that does the fourth and skips the first three has a beautifully written text file describing a website nothing can read.

Questions we get asked

Will this get me into ChatGPT?

On its own, no, and be wary of anyone who says otherwise. What gets a business named by an assistant is the ordinary stuff: content it can read, facts it can verify, consistency across sources, and being mentioned by other websites it already trusts. llms.txt is a small, cheap addition to that, not a substitute for it.

Is there a downside?

One, and it is worth taking seriously: it is a public statement about your business, so it needs to be accurate. If it claims services you dropped or an address you moved out of, you have created a tidy, machine-readable source of wrong information about yourself. Treat it like a page, not like a config file.

Do I need one if I already have good schema markup?

Not urgently. Structured data is doing the heavy lifting and it is the one with documented effects today. Add llms.txt when it is convenient, not instead of anything.

How do I know if I have one?

Type your domain followed by /llms.txt into a browser. If you get a 404, you do not have one. Worth knowing: most agency websites on this island do not, including some that sell AI visibility as a service.

If you want a broader picture than a single file, our free checker reads your site the way an assistant does and tells you what it could and could not find — structured data, readable content, speed and all. About a minute, no signup.