Authorial voice
Authorial voice is the consistent personality of the prose itself: rhythm, diction, and attitude that make a book sound like one writer wrote it. Here is what it is and why AI tends to flatten it.
Authorial voice is the personality of the prose. It is the sum of your sentence rhythm, your word choices, the things you notice, and the attitude underneath it all. Two writers can describe the same room and produce completely different pages, because voice is the part that is unmistakably theirs.
Voice works at two levels in fiction. There is the author's voice, the consistent style across the whole book, and there are character voices, the distinct way each person speaks and thinks. Strong writing holds both at once: the narration sounds like you, and the dialogue sounds like the people.
Why AI flattens voice
Most language models default to a smooth, agreeable, slightly corporate register. Left alone, they sand the edges off your prose until every paragraph reads like every other paragraph on the internet. The result is fluent and forgettable, the opposite of voice.
The fix is not a single setting. It is constant comparison against samples of how the writing should actually sound, plus a refusal to let the model drift back toward its default. A few signs your voice has been flattened:
- Sentences that are all the same length and shape.
- Safe, generic word choices where a specific one belonged.
- Em-dashes and hedging phrases that no human chose.
- Description that could describe anything.
How FireQuill protects voice
FireQuill learns a character's voice from real samples and checks new dialogue against it, flagging the moment a line drifts off-key. The same discipline runs on the prose level, so generated chapters keep sounding like the book and not like a model. The goal is the thing the platform is named for: writing that does not sound like AI wrote it. See AI slop for what that failure looks like up close.