Does one-click AI rewrite actually work? How PaperMirror retests

“Does one-click AI rewrite work?” splits communities in two: sellers promise 100% wipe, buyers say it got redder. They are not talking about the same thing. Sellers declare victory on their own detector. Buyers get hit by the school report.

When it really is useless

• The whole file goes into a black box. Nobody looks at which paragraph is exploding.
• Only synonyms change. The syntax skeleton stays.
• Success is judged on a completely different model.
• The output is unreadable. An instructor does not need a detector.

Why PaperMirror can say rewrite is not a prop

1. Same internal engine: MirrorEngine both scores and accepts the rewrite. A drop is not a changed ruler.
2. Paragraph surgery: high-risk bands first. Your structure and citations stay.
3. Immediate retest: new text returns to the report. Overall rate and three colours update.
4. Price matches cost: rewrite is 2 credits/character because a generation model actually runs here. Detection is cheaper so you can iterate.

In internal evals, humanizing rewrite pushed many high-risk Chinese samples into the low-risk band. English also drops, with magnitude depending on the detector. That is why the product keeps retest instead of declaring victory after one click.

Will the next model see through the rewrite?

Detectors iterate. We sell a repeatable loop, not a permanent charm. A scan the night before submit is always more useful than a screenshot from three weeks ago.