Most people prompt AI every time they need something. There is a different model — configuration over prompting — where the AI already knows your voice, your audience, and your system before you type a word.
Most advice on writing a Claude voice prompt is too vague to be useful. Here is what actually belongs in one, with real examples from an actual voice profile rather than hypothetical ones.
Most people use “write in my voice” as a tone instruction. Claude needs a lot more than that — and the gap between what you said and what it actually needs is where the fabricated anecdotes come from.
Most people calculate whether AI saves them time. Almost nobody calculates how much of their editing time is just correcting AI output back toward their own voice. That part is fixable.
The problem with AI content isn’t that the AI is bad at writing. It’s that it’s writing from a generic baseline, and you’re spending all your editing time correcting back toward your actual voice. Here’s what fixes it.