How to lower an AI score: scan, rewrite the red band, retest
Direct answer The right way to lower a paper’s AI score is: scan the full draft on one engine, rewrite only high-risk red paragraphs, retest immediately, then read it yourself. PaperMirror puts those steps on one report and bills credits per character. We cannot guarantee a school-system score and we do not ghostwrite.
TL;DR Lowering an AI score is not random synonym swapping. The order is: scan the whole paper → edit only the red band → retest → read it yourself. PaperMirror puts those three steps on one report, billed per character.
Search “how to lower AI score” and you get either a manual conjunction swap, or a humanizer link. The first is slow. The second sends you to another site to re-scan. PaperMirror follows the closed loop students already expect: detect → process → re-check → export. The reading we align with is a Turnitin-style report, not nine domestic Chinese platforms.
Step 1: Scan a full paper or chapter
- Paste text or upload .txt / .docx, up to 20,000 characters per run.
- Do not scan under 200 characters — the score will drift.
- Signed-in reports show high / medium / low colour. Guest trials only show the overall rate and a risk-paragraph count.
- Detection is 1 credit = 1 character. One free scan up to 1,000 characters per day.
Step 2: Hit high-risk paragraphs only
Washing the whole file scatters the argument and breaks terms and citations. Edit against the red paragraphs: swap in your data, interviews, failed experiments, class discussion — or use PaperMirror one-click rewrite. Rewrite is 2 credits / character, so it is more expensive than a scan. That is why you should only rewrite what needs it.
Step 3: Retest in place
Score it again with the same MirrorEngine. Watch two things: whether the overall AI rate enters the low-risk band, and whether continuous red paragraphs disappear. Medium-risk bands can wait for the next round. Do not rewrite on site A and judge success on site B. Two algorithms will spin you.
Step 4: Read it yourself, then submit
- Proper names, scale names, and statute names were not mangled.
- Citation years and page numbers are still there.
- Paragraphs still connect.
- Figure captions still match the body.
| Action | Credits | About CNY (¥1 = 1,000 credits) |
|---|---|---|
| One full scan | 8,000 | ¥8 |
| Rewrite 2,000 red characters | 4,000 | ¥8 |
| One retest | 8,000 | ¥8 |
| Another 500-character edit + retest | 1,000+8,000 | about ¥9 |
A serious round is a little over ¥20. The ¥30 standard pack is more comfortable after bonus credits. Dissertations that cycle many times belong on the ¥50 / ¥100 tiers. Credits never expire.
Run the three steps on your draft
Try a scan first, then sign in for the full report. Red paragraphs can be rewritten in place.
Scan and rewriteWhen the score will not drop, it is usually one of these
- You only swapped synonyms. The paragraph skeleton is still the model’s.
- Methods / discussion have none of your own operating detail.
- The text is too short, or you only pasted the introduction.
- You rewrote on another site, then used a different detector to decide “whether it worked.”
FAQ
- Do I have to get to 0%?
- No. 0% is neither realistic nor healthy. See what AI rate is normal: low-risk band plus no long highlight streaks is enough.
- Does this work on Chinese papers too?
- Yes. PaperMirror detects Chinese and English automatically. If your domestic final reading is CNKI (China’s academic database / AIGC checker), the school system is still the source of truth. PaperMirror is for clearing generative paragraphs before you submit.
Detection and rewrite are for pre-submit self-checks. Follow your school's integrity policy. PaperMirror is not Turnitin or CNKI. The school system is the source of truth.