Writer proposes 'Who Wrote This?' AI-use scale

The proposal arrives after a loaded episode in which billionaire investor Stanley Druckenmiller admitted using AI to write a Wall Street Journal op-ed attacking Treasury Secretary Scott Bessent's bond-market interventions; the AI detector Pangram scored the piece 100 percent A.I., even as Druckenmiller said he rejected many suggestions and AI didn't write 'the whole thing.' The Journal's editorial page editor, Paul Gigot, was unapologetic, calling AI 'a fact of modern life,' and later writing that Druckenmiller's sin was admitting its use. The author argues the binary 'Did you use A.I.?' question cannot capture that variation, and that writing lacks math's single right answer.

The commentator writes from inside the news-credibility world: he tracked AI-generated misinformation at NewsGuard and built an AI propaganda site for the Wall Street Journal for $105, and now runs Newsreel, a media startup that vets independent journalists before they publish. Even with that vetting, he says readers still want to know how much AI went into what they read. He notes writers already fear an AI witch hunt — some scrubbing em dashes, and in April The Journal's James Taranto compared Pangram to a 'defamation machine' — and reaches back to an 1845 accusation that Alexandre Dumas ran a 'novel factory' to argue the 'Who wrote this?' question never disappeared. He discloses his own piece as H3 Polished.

The stakes appear to hinge on whether self-reporting holds up and whether publishers adopt a label in the first place, since the scale only works if readers trust the disclosure; the author suggests detector exposure could keep most writers honest, but he concedes some may be tempted to shade the truth. For readers, the payoff is an answer without running every piece through a detector — but only if enough outlets print it.