Who Can You Hide Behind? Author-Conditioned Style Transfer and the Evasion of a Commercial AI-Text Detector
Organizations: Varendra University, Rajshahi, Bangladesh · University of Rajshahi, Rajshahi, Bangladesh
Abstract
AI-text detectors can be sensitive to changes in writing style. We examined this sensitivity by rewriting 991 AI-generated answers with Gemini, using prompts that included a writing sample from one of 55 named authors. We scored each original and its rewrite with the QuillBot AI detector. The median score fell from 100% to 0%, and the mean fell from 67.9% to 11.2%. Of the rewrites, 66% received a score of exactly 0%. Among the 699 originals scored at least 50%, 71% of rewrites scored below 10% (paired Wilcoxon signed-rank test, p-value below 1e-100). Scores varied across the authors used in the prompts. All rewrites associated with 10 authors scored 0%, whereas the mean score remained above 50% for six authors (Kruskal-Wallis test, p-value below 1e-80). Greater reuse of five-word sequences from the author sample was associated with lower scores, although rewrites with less than 5% reuse still averaged 15.2%. These results describe one detector and one rewriting system, not detectors in general. The study lacks a generic-rewrite baseline, did not measure preservation of meaning, and produced rewrites that were substantially longer than the originals.
Figures & tables
| Subset | Before | After | ||||
| All pairs | 991 | 67.9% | 11.2% | 653 (66%) | 708 (71%) | 82 (8%) |
| Orig. | 699 | 95.1% | 12.0% | 454 (65%) | 494 (71%) | 67 (10%) |
| Author | At 0% | Reuse | Len. | |
| Adam Grant | 14 | 7% | 0% | 6.4 |
| Agnes Callard | 16 | 31% | 2% | 4.4 |
| Ali Abdaal | 6 | 17% | 1% | 6.2 |
| Anderson Cooper | 19 | 16% | 0% | 2.7 |
| Andrej Karpathy | 17 | 24% | 8% | 5.9 |
| Anne Applebaum | 21 | 67% | 5% | 3.5 |
| 5-grams reused | Mean after | After | |
| Under 5% | 663 | 15.2% | 376 (57%) |
| 5% to 20% | 213 | 3.9% | 167 (78%) |
| Over 20% | 115 | 1.5% | 110 (96%) |
Appendix figures & tables1 asset
Supplementary material from the paper’s appendix.
Appendix
| Sheet row | Problem | Action |
| 99 rows | Empty | Dropped |
| 100, 192, 290, 918, 1000 | Rewrite score missing | Dropped |
| 110 | Score cell holds pasted text, no author | Dropped |
| 535, 902 | No author name | Dropped |
| 904 | Original score recorded as 1000% | Dropped |
| 222–240 | Author spelled “George Monbio” | Reported as George Monbiot |
Robustness of Machine-Generated Text Detection
Robustness of machine-generated text detection studies whether AI-text detectors remain reliable across shifts, edits, languages, and adversarial attacks.
See all 103 papers