AI Detection Bypass Tools Tested : Only One Scored 0% on Turnitin
TL;DR
Andy Stapleton tested three AI-detection bypass methods against Turnitin and Originality.ai.
Key Points
- Only one tool – 'Humanize' – achieved a 0% detection rate specifically on Turnitin.
- The other two methods produced inconsistent results, remaining detectable by at least one system.
- AI detectors are reportedly getting sharper, meaning simple paraphrasing tools increasingly fall short.
Nauti's Take
A tool that reliably fools Turnitin at 0% is not a curiosity – it is a structural problem for academic integrity systems. What is more striking than the result itself is the speed at which such bypasses get publicly documented and distributed.
Anyone who believes detection alone solves the problem is missing the point: the real question is whether text-submission-based assessments are still fit for purpose in an age of capable AI writing tools.
Context
The test illustrates the escalating arms race between AI generation and AI detection, with real consequences for academic institutions, publishers, and anyone relying on text authenticity. A single tool consistently hitting 0% detection becomes a de-facto cheat standard, forcing platforms like Turnitin into another round of model updates. At the same time, the results confirm that no detector should be treated as infallible – a dangerous signal for those who rely on them uncritically.
Video
Sources