Ghostwriting and AI detection — an awkward combination
Ghostwriting was already an exercise in writing like someone you’re not. A detector measures pattern, not authorship — which makes it a strange tool to point at this particular job.
What a detector can and can’t tell a client
It can say whether a piece reads as statistically typical of AI-generated text. It cannot say who typed it — and for ghostwriting specifically, “who typed it” was never supposed to be the byline author anyway. That’s the whole arrangement.
Where it’s actually useful in this work
As a check on the generation tools some ghostwriters use as a starting point, before rewriting fully in the client’s voice. If a draft still reads as templated after a real rewrite pass, that’s useful editorial signal regardless of where the first draft came from.
Set the expectation with the client up front
If a client’s own brief or contract requires disclosure of AI tool use, that requirement doesn’t disappear because the byline was already someone else’s. Decide, and say plainly, whether tools were used at any stage — see what you can honestly promise a client for language that doesn’t overstate what a check proves.
The genre problem
Ghostwritten executive bios, LinkedIn posts and press quotes are often deliberately formulaic — that’s part of the job. Formulaic writing reads as templated to a statistical detector independent of how it was produced, so a mid-range score on this kind of piece is weak evidence either way.