AI Won't Copy Your Voice. That's a Good Thing.

OpenAI quietly updated ChatGPT in August to refuse requests to copy the style of specific named authors. "Write this in the style of Stephen King" now gets declined.

The AI creator community is split. Half the people I talk to think it's censorship. The other half think it doesn't go far enough. I think it's the first step toward something that will make original creators more valuable, not less.

Here's my reasoning. The fundamental promise of AI-assisted creation is that it amplifies what you bring to the table. Your expertise. Your perspective. Your hard-won knowledge from years of actually doing the thing. That's the argument I make in my book, and it's the argument I believe in. But style mimicry undercuts that promise. It says: you don't need your own voice. Just borrow someone else's.

That was always a dead end. A generated imitation of someone's style is a copy of a copy. It captures the surface patterns — the sentence rhythms, the vocabulary range, the structural habits — without any of the substance that made those patterns matter in the first place. King writes the way he does because of what he's seen and thought and lived. The style is an output, not an input. Copying the output without the substrate produces something that reads like King to someone who's never read King.

For builders and creators who have developed their own voice — through practice, through failure, through years of doing the work — this policy is protective. It means AI becomes a tool that helps you express your voice more effectively, not a tool that lets anyone impersonate it.

I'm writing these posts in my voice. It took me two books and a year of public writing to find it. I don't want AI to be able to reproduce it on demand for someone who hasn't done that work. Not because I'm precious about it — because the voice is the value. Take that away, and there's nothing left worth reading.