Amazon grades your book for AI, and the one real method flags the realest humans

An alarmist post making the rounds claims Amazon secretly detects AI in Kindle books: keystroke rhythms, invisible Unicode timing marks, revision-history forensics, cryptographic proof you sweated over the draft. Most of it is invented. First off, almost nobody writes inside Amazon’s editor, so there is no keystroke stream to read; a finished upload carries none of this.

And there’s no source under it. Every page that “confirms” the keystroke forensics is another SEO post echoing the same couple of phrases — “digital DNA,” “fourteen minutes of editing in the XML” — back to the one before it. No Amazon documentation, no reporting, no patent that survives a look. A fair number of them read like they were written by the machines they warn you about. An unverified claim, photocopied sideways until the echo passed for proof.

One item on the list is real. Stylometry — matching a book’s voice against a known fingerprint. And that’s the one worth thinking about, though not for the reason the post gives.

Stylometry rewards consistency. A distinct voice held steady across a body of work is exactly what it grabs hardest — and a distinct, steady voice is also the thing a real writer spends years earning. I once ran my own pre-AI archive through a detector and watched it call 2008 me a machine. The tells were mine first .

So look at the shape of it. The gate that used to ask whether you were allowed in now asks whether you’re real. It decides using rules it won’t publish, and its sharpest instrument indicts the people it claims to protect. We have watched an industry try to draw this exact line before, under better conditions, and it fell on the wrong people .