AI Detectors Fail When Imitating Writing Styles
FlorianGallwitz · x · 2026-07-18
An Epoch AI study tested several AI detectors across different text types:
- For human writing, detectors rarely misidentified it as AI text.
- For directly generated LLM text, the false negative rate was extremely low, around 0%0.7%.
- However, when an LLM was prompted to mimic a real author's style, detectors failed noticeably: the false negative rate across different tools jumped to roughly 10%18%.
- In scientific writing scenarios, the false negative rate was even higher, reaching about 26%.
The chart also details sample sizes and error margins. The conclusion is that while AI detectors are good at catching "blatant AI text," their reliability drops significantly when faced with content designed to mimic specific authors.
More from Research
- Cognition's SWE-2 uses a KKT duality argument in RL to shift the effort Pareto curve — YouJiacheng · 2026-09-11
- VidMap uses RoMa coarse matching on all frames, fine-scale only for keyframes — ducha_aiki · 2026-09-11
- Bug Hunt Bench author: leaderboard noise is about 2-3 points — PawelHuryn · 2026-09-11
- PNAS paper shows a tiny billiard-ball system is a universal computer — undecidability lives in two dimensions — eigensteve · 2026-09-11
- New paper: Absolute pose estimation from affine cues and gravity direction — ducha_aiki · 2026-09-11
- LoMa Paper Ships REALLY HardPairs Dataset, Accepted at ECCV 2026 — ducha_aiki · 2026-09-11