Over-detect hallucinate AI contents
Your AI detection tool is producing highly questionable results and appears to be fundamentally unreliable. Even very short and simple sentences writing are almost entirely flagged as AI-generated, and when I tested published academic papers, the reported AI percentages were even higher and most of time 100%, which is clearly unreasonable. For a system used in academic integrity decisions, this level of over-detection is unacceptable and risks serious harm to students and researchers. The algorithm currently seems overly sensitive and not fit for its purpose, and it requires immediate review and correction.