The Detection Deception, Part 2: Publishing's Trust Problem Isn't AI — It's Process

In Part 1, John Ayers made the case that AI detectors don't work. The math is against them, the research is against them, and OpenAI couldn't even build one that was reliable. Part 2 picks up where that left off: what happens when the publishing industry uses these tools anyway — and people get hurt.

Drawing on reporting from Kathleen Schmidt's Publishing Confidential, Ayers walks through three cases that, taken together, expose an industry with no infrastructure for verifying authenticity. Hachette pulled Mia Ballard's horror novel Shy Girl after an AI detector returned an 84% score — the same class of tool that flags Schmidt's own writing at 90%. Amy Griffin's Oprah Book Club memoir is now the subject of a lawsuit because no one verified whether the central traumatic memory actually belonged to her. Woody Brown's bestseller raised authorship questions that Penguin Random House's Hogarth imprint declined to address.

There is no fact-checking in book publishing. Publishers do not have fact-checking departments. — Kathleen Schmidt

Three different publishers. Three different genres. The same underlying failure: no system for verifying that the person on the cover actually has the expertise, the story, or the authorship the book claims. Bolting AI detectors onto that vacuum doesn't fix it — it just adds a new way to wrongly destroy a writer's career.

Ayers argues the industry is chasing the wrong question. "Did AI touch this?" is unanswerable and increasingly meaningless. "Is the expertise behind this content real, and can we prove it?" is the question that matters — and it requires process, not scanning. At Chapters, that process starts with a real human, a senior editor, recorded interviews, transcription and organization assisted by AI, and a documented chain of review and approval at every stage. Provenance, not perplexity.