What is llms.txt — and should your website have one?

A one-page markdown map of your website, written for language models instead of browsers. Cheap to add, zero risk — and surrounded by more hype than facts. Here is the honest version.

The idea

llms.txt is a proposal (llmstxt.org, September 2024) for a plain markdown file at the root of your domain — example.com/llms.txt — that gives language models a curated summary of what your site contains and where the important pages are. Where robots.txt controls access, llms.txt describes content: think of it as a table of contents written for machines with small attention spans.

The format is deliberately trivial — a heading, a summary, and link lists:

# Example Plumbing AS

> Plumbing services for homes and businesses in Oslo — emergency callouts, bathroom renovation, heat pumps. Established 2005, 12 employees.

## Services
- [Emergency plumber](https://example.com/emergency): 24/7 callout in Oslo
- [Bathroom renovation](https://example.com/bathroom): fixed-price projects

## Company
- [About us](https://example.com/about): history, team, certifications
- [Contact](https://example.com/contact): phone, email, address, org.nr

The honest status (July 2026)

No major AI provider has officially confirmed consuming llms.txt. Not OpenAI, not Anthropic, not Google, not Perplexity. Anyone selling llms.txt as a guaranteed visibility boost is ahead of the evidence — and plenty are selling exactly that.

So why bother? Three reasons. It costs fifteen minutes and carries zero risk — no crawler is harmed by an extra text file. A growing set of developer tools and smaller AI agents do read it today. And writing one forces the useful exercise of deciding, in one page, what your business actually is and which pages matter — a summary that also tends to expose what your website itself fails to say clearly.

That reasoning is reflected in our audit methodology: llms.txt earns a single bonus point — rewarded if present, never punished if missing. That is the proportionate weight for an emerging standard.

Almost nobody has one

In our benchmark of 447 Norwegian trade businesses, 7% have an llms.txt. For comparison: 92% have robots.txt and 87% have a sitemap. If the standard does get adopted by major providers, the businesses that already ship it start ahead; if it never does, they lost fifteen minutes.

Writing yours in 15 minutes

Start with one sentence a stranger could act on: who you are, what you do, where. Add your services as a link list with a half-line description each, then company pages (about, contact — include your organisation number). Keep it under a page, in the language of your customers, and update it when your offering changes. Ours is live at appsynth.pl/llms.txt if you want a real example.

One warning: llms.txt is a complement, not a shortcut. If your actual pages are invisible to AI — blocked bots, JavaScript-only content, no structured data — a beautiful llms.txt fixes nothing. Get the fundamentals right first.

Where do you stand on the fundamentals?

The free mini-audit runs 5 of our 33 checks on your domain — the ones that decide whether AI can read and understand your website at all.

Get a free mini-audit → Guide: robots.txt for AI