New tool detects fake, AI-produced scientific articles

Ahmed Abdeen Hamed, a researcher at Binghamton University, developed xFakeSci, an algorithm that detects 94% of AI-generated fake scientific papers, especially in biomedical fields. Collaborating with Xindong Wu, Hamed compared 50 AI-generated articles on Alzheimer’s, cancer, and depression with real ones, finding fewer bigrams (common word pairs) and more interconnected language in fake papers. He highlights that AI like ChatGPT focuses on convincing arguments, while real science emphasizes honest reporting. Despite its success, xFakeSci still misses 6% of fakes, prompting further research to improve detection across various fields.