AI Detection Tools in Higher Education: Evidence, Risks, and Recommendations for Institutional Practice

AI detection tools purport to identify whether student writing was produced by generative AI systems. Current research, however, documents substantial limitations, including low reliability across contexts, higher misclassification rates for multilingual and machine-translated writing, vulnerability to circumvention, opacity, and risks to student well-being and due process. While detectors may occasionally prompt reflection on writing processes, their limitations significantly constrain the role they can responsibly play in academic integrity decision-making. For these reasons, the committee does not find sufficient value added to justify institutional adoption of AI detection tools or the use of detector outputs as evidence of misconduct. Institutions can better support academic integrity by investing in assignment design, transparency around AI use, and developmental conversations with students about their learning processes.