2026-08-01 · Jack Stovell

llms.txt for people: what it is, and an honest take on whether it matters

llms.txt is a proposed convention: a markdown file at the root of your site that tells language models what the site contains and where the good stuff is, the way robots.txt tells crawlers where they may go and sitemap.xml tells them what exists.

You'll see confident claims in both directions about it. Let's be precise about what's actually known.

The honest status

That combination — unproven upside, near-zero cost, no downside — is why hilyt ships it everywhere rather than waiting for the argument to resolve.

From site index to identity card

The more interesting move is applying the same idea to a person or a business instead of a site. That's what a hilyt llm.txt card is: one plain-text page per profile that states, in a shape a model digests losslessly, who someone is.

A good card reads like a structured dossier, not marketing copy:

The point isn't the file format. The point is that when an AI retrieves it, there is nothing to misread: no navigation, no cookie banner, no JavaScript, no ambiguity about which entity the page describes.

Do models actually read these?

Retrieval-based tools read whatever plain, open text ranks or gets linked next to your name. A text card at a stable URL, linked from your other profiles, is exactly the kind of page retrieval rewards. That part isn't speculative — it's how retrieval works today.

Whether the llms.txt convention specifically wins is genuinely open. If it does, you're early. If it doesn't, you still own a clean, crawlable, disambiguated statement of record. Either way you're better off than a login-walled profile.

Every hilyt profile gets one automatically at i.hilyt.it/yourname/llm.txt, kept in step with the human-readable page.

See a live llm.txt card