What Is llms.txt?

In one sentence

llms.txt is a proposed convention: a markdown file at your domain root summarising your site's key content for language models.

What llms.txt means

The idea is a curated, machine-friendly index — a short description of the site plus links to its most important pages, in plain markdown rather than HTML.

It's a community proposal rather than a standard from any major platform, and adoption by the systems it targets remains genuinely uncertain.

Why it matters for AI search

We implement it because it takes fifteen minutes, costs nothing, and might become meaningful. We won't pretend it's more than that.

Any agency positioning llms.txt as a core service line is dressing up a trivial task. The things that actually decide AI visibility are crawler access, rendering, entity clarity and corroboration.

See it in action

What an llms.txt file actually contains.

A working llms.txt

File structure
// click generate

Cheap to implement, uncertain to matter.

Treat it as a hedge. Fifteen minutes of work for a possible future benefit is a reasonable trade — building a strategy around it is not.

How to get it right

Implementing it sensibly

  • Put a plain markdown file at yourdomain.com/llms.txt
  • Summarise what the organisation does in one or two sentences
  • Link the pages you'd most want cited, grouped by section
  • Keep it current — a stale file is worse than none
  • Spend the remaining effort on robots.txt, rendering and entity work, which demonstrably matter

Common questions

Does llms.txt actually do anything?

Adoption by major AI platforms is unconfirmed. Some tools read it; the systems that matter most haven't committed to it.

We implement it as a cheap hedge, not because we can show it moves anything.

Is it a replacement for robots.txt?

No. robots.txt controls crawler access and is respected by every compliant crawler. llms.txt is descriptive and optional.

If you only do one, do robots.txt properly.

Should we pay an agency for llms.txt?

Not as a line item. It's a fifteen-minute task that should be included in any technical engagement.

If it's being sold as a distinct service, that's a signal about the rest of the offering.

These come up alongside llms.txt constantly.

Free 12-point check

Send Ron your site

Two fields. We’ll crawl your site as GPTBot, ClaudeBot and PerplexityBot, benchmark a sample of your category’s prompts, and send you what we find.

  • What each AI crawler actually receives from your site
  • Whether models resolve your brand as a real entity
  • A sample of category prompts and who gets cited
  • The three fixes we’d prioritise first

No commitment, no sales sequence. If we’re not a fit, we’ll say so.

We’ll pull your favicon so you know we found the right site.

We reply from a real address. No sequence, no newsletter.