Using AI Tools Without Losing Your Voice

The temptation with a generative model is to let it finish your sentences. The discipline is to let it surface the sentence you would have written if you had more time, more sleep, and a better reference library. The model is not your voice. It is your assistant in a hurry.

Use models for friction, not for fluency. Ask them to argue with your draft. Ask them for the version a sceptic would write. Ask them what is missing, what is repeated, what is condescending, what is unclear to a fifteen-year-old. Then close the tab and write the paragraph yourself. The model gave you a map. You still have to walk.

Beware of the average. A large language model is, by design, statistically average. If you accept its first draft, you will sound like the middle of the internet. Your job is to push past the middle. Read what it gave you, then write the version that sounds like you on your best day, the day you remembered the exact word, the exact verse, the exact moment.

Protect your inputs. The reason you have a voice in the first place is that you have been formed by specific books, specific people, specific seasons of suffering and joy. Keep reading those books. Keep talking to those people. A creator who only consumes model output will eventually only produce model output. Garbage in, garbage out is a software cliché. It is also a discipleship principle.

Cite generously. When a model gave you an idea worth keeping, say so. When a real human shaped your thinking, say so even louder. The future belongs to creators who are transparent about their process, because trust will be the scarcest resource on the internet.

Your voice is the asset. The model is the editor in a hurry. Never trade those roles.