I got tired of AI systems citing my drafts as if I meant them — so I wrote a spec

Last year I published a rough draft, forgot about it, and months later found it cited in an AI-generated summary as if it were my considered opinion. The page had no way to say "this was never finished."

robots.txt has been solving the wrong problem for 30 years. It controls access — not intent. There's no standard way to mark a page as draft, temporary, or anonymous before a crawler decides what to do with it.

So I wrote one.

Untitled Protocol is a single meta tag or HTTP header:

<meta name="untitled" content="draft">
<meta name="untitled" content="ephemeral,anonymous">

Six states cover most cases: now, draft, ephemeral, anonymous, collective, generative. Compliant agents — crawlers, AI pipelines, archivers — can read this and treat the content accordingly.

I also built a live validator, an AI suggestion tool that reads a URL and recommends which state fits, a 1kb JS widget, an npm package, and a WordPress plugin.

The spec is MIT licensed. I want it to become a real standard — not my product.

Would love to know: has anyone else run into this problem, or found a workaround I missed?