What is llms.txt?A plain answer, and a real one.
An llms.txt is a text file at the root of your site that tells a language model what your business is, what it sells and which pages matter. It was proposed in September 2024 and it is not a standard anybody has ratified. No search or AI company has said it affects your rankings. It is still worth publishing, and this page explains exactly why, with ours as the example.
Four parts, in this order.
The format is Markdown and the whole thing is usually under a hundred lines. What makes one good is not length, it is what you left out.
An H1 with the business name
Nothing else. This is the title a model quotes when it names you, so it should be the name you trade under, not a tagline and not a keyword string.
One blockquote that says what you do
A single sentence covering what you sell, who for, and where. If a model only reads one line of the file, this is the one it reads, and it is the sentence that ends up inside an answer about you.
Sections for what you actually sell
One heading per product or service, with the facts under it: what it includes, who it is for, what it costs if you publish prices, which areas you cover. Plain sentences beat marketing copy here, because the reader is a machine summarising you to somebody else.
The pages that matter, and nothing else
A short list of the URLs you would want quoted. Leaving out the thin pages is the point: the file is a curated map, and its value comes from what you chose not to include.
It is not a second robots.txt.
One is a gate and the other is a briefing. They do different jobs, they do not replace each other, and you want both.
Nobody has confirmed it moves rankings. Publish one anyway.
Google has not said it reads llms.txt. Neither has OpenAI, Anthropic or Perplexity. Any page telling you it is a ranking factor is selling you something on a guess, and most of the pages ranking for this term are doing exactly that.
Here is the case that survives that. Models already describe your business, whether or not you help them. Without a file they assemble that description from your navigation, your footer and whatever a directory wrote about you three years ago. With one, there is a current, accurate paragraph in your own words at an address anything can find. The file costs an afternoon. Being wrongly described in the answer a customer reads costs considerably more.
That is the whole argument. It is smaller than the one you will read elsewhere, and it is the one that will still be true next year.
They dropped their ad agencyand kept our system.
A multi-location local service business, running in three markets. We moved them onto a new site with nothing on it and opened a market they had never appeared in at all, so both started from actual zero. A year later they could see exactly where their leads came from, and it was not the paid ads.
Their third market had no presence at all when we started. The percentages are the share of tracked queries where they get named. Most competitors are named on none.
What people ask about llms.txt.
Let's see how AI assistants describe your business today.
Book a free consultation. Thirty minutes on what the engines currently say about you, which of it is wrong, and what the engine would fix first.
