I write this with some trepidation, given that I make regular use of AI tools in my own writing process. The temptation toward hypocrisy is high.
Let me try to be precise about what I mean.
The voice problem
A voice in writing is the accumulation of a thousand small decisions: word choice, sentence rhythm, when to use a comma and when to use a dash, how long you’ll let a thought breathe before landing it. It takes years to develop and it’s deeply personal.
AI writing tools, even good ones, tend toward a kind of averaged competence. The sentences are correct. The structure is logical. The argument proceeds in orderly fashion. What’s missing is the idiosyncratic move — the unexpected comparison, the sentence that goes longer than it should because the thought requires it.
Where it actually helps
Structure. I’ll often use a model to check my argument’s bones — does the logic hold? Is there an obvious objection I haven’t addressed? This is more like having a rigorous editor than having a ghostwriter.
Research breadth. What am I not thinking about here? is a genuinely useful prompt. The model can surface adjacent ideas that I might not have reached through my own reading.
Editing for clarity. Not for style — for clarity. Is this sentence doing what I think it’s doing? Sometimes the answer is illuminating.
The line
Here’s where I land: if you’re using AI to generate the sentences that go into your published work, you’re not writing — you’re curating. That might be fine, depending on what you’re trying to do. But you should be honest with yourself about which one it is.
The voice is yours or it isn’t. The tool can help you find it more efficiently. It can’t be it for you.