Does my business need an llms.txt file?

Not today, though it costs little to keep. An llms.txt file is a plain-text summary of your business, placed at yourdomain.com/llms.txt, written for AI models to read. As of September 2026, OpenAI, Anthropic, Google and Perplexity have not documented reading it, and Ahrefs found 97% of llms.txt files got zero requests in May 2026 (137,210 domains, published 15 Jun 2026).

That is the short version. But "do I need one" is the wrong first question. The better question is whether AI can clearly understand your business today, because that problem is solved on your site, your profiles and third-party sources, the pages crawlers actually fetch. An llms.txt file only restates those facts. Google said in June 2026 that it is not needed for Google Search, and John Mueller called it "purely speculative for now" (2 Jun 2026).

What an llms.txt file actually is

It is a Markdown text file served at the root of your domain, reachable at yourdomain.com/llms.txt, next to robots.txt. Inside, it states the key facts about your business in a form a model can lift directly: who you are, what you do, who you serve, your proof, and where to verify it. robots.txt is a standard crawlers obey. llms.txt is a proposed format that no major AI platform has said it reads, as of September 2026.

It is not magic and it is not a ranking hack. It is the difference between making a model assemble a blurry picture of you from scattered, inconsistent web content, and handing it a clean statement it can trust and repeat.

So do you need one?

Run a quick test before deciding. Ask ChatGPT or Perplexity to describe your business, and ask them who they would recommend in your category. Then read the answers honestly. You will land in one of three places.

• The model describes you wrong, or vaguely. It is guessing, because the web does not give it a clear picture. Fix the sources first: your site, About page and profiles. An llms.txt file will not correct it on its own.
• The model ignores you and names competitors. You are not legible enough to be in the answer at all. The file helps, as part of the broader work of being findable and trusted.
• The model describes you accurately and recommends you. You are already legible. The file is polish and insurance, worth doing, but you are not bleeding.

Most businesses are in the first two. If you are, the work is on the pages and profiles models actually read. The file is a tidy copy of that work, not the fix.

What a strong one contains

Here is where most llms.txt files fail. They exist, but they are a thin paragraph that adds nothing. A weak file is a name and a sentence. A strong file is a structured profile that answers the real questions a buyer would ask an AI, so the model can pull the right answer for each one.

The sections that matter: a summary that states who you are using the terms a model matches on, a differentiation section with a claim a competitor cannot honestly make, a clear definition of who you are for, proof that includes the mechanism and a way to verify it, an embedded FAQ answering the exact questions buyers ask, authority signals that are hard to fake, and contact details with links the model can follow to confirm everything. The thin version skips all of this. The strong version is a useful fact sheet for you; there is no public evidence yet that any model reads it.

The mistake to avoid

The biggest mistake is publishing an llms.txt file whose facts do not match the rest of your site and the web. Models cross-check. If your file claims something your site contradicts, you do not gain trust, you lose it, and a model that catches an inconsistency hedges instead of citing. An llms.txt file is not a place to inflate. It is a place to state, clearly and verifiably, what is already true and provable elsewhere. Consistency is the whole point.