You can tell when something was written by AI. So can your readers.
Not because it is badly written. Because it sounds like nobody. That is not the model failing — it has never met you, so it hands you the average of everyone who has ever written about your topic.
The Generic Detector
One prompt. Paste any draft into it and it scores the writing 0–100 on how anonymous it reads, quotes the exact sentences dragging the number up, and rewrites them in a voice that carries a signature.
It arrives immediately, as plain text in the email — no PDF, nothing to download. Roughly one email a month after that, and the unsubscribe link works on the first click.
A score you can argue with.
What the number means0–20Unmistakably one specific person
21–40Mostly distinctive, some filler
41–60Competent and forgettable
61–80Could be any brand account
81–100Indistinguishable from default model output
It quotes your own sentences back at you.
Most feedback on AI writing stops at "add more personality." This names the sentence, names the pattern it matches, and shows the fix at the same length. Five sections, every time.
- Anonymity score — the number, 0 to 100
- Verdict — one blunt sentence
- The tells — every phrase dragging the score up, quoted, with a rewrite
- What is already working — the lines to protect, not just the ones to cut
- The one fix — if you change nothing else, change this