AI detection software does not work. Period. A recent study from academic researchers highlights that these tools are extremely unreliable. When simple anti-detection techniques are used, their limited accuracy further deteriorates.
This implies that when these tools identify AI cheating, they are only catching poorly executed attempts. Educational institutions have relied on plagiarism detection tools for internet research and Wikipedia content. It seems logical to seek similar solutions for GenAI, but these tools are simply not fit for purpose and can even cause harm, as seen in the case of a UNG student placed on academic probation for using Grammarly.
The future does not lie in these detectors. Instead, we need to shift our conversations about academic integrity to address how assessment must evolve in the age of GenAI.
You can read more about the study here.
