A Stanford University study found that up to 17% of peer reviews in major computer-science publications are now AI-generated. The research analysed 50,000 peer reviews from recent years, identifying AI-generated reviews by their formal tone, verbosity, and superficial content. The use of AI for peer reviews is attributed to time constraints. The study raises concerns about scientific integrity and AI limitations in providing in-depth reasoning. Recommendations include establishing guidelines for responsible AI use in peer reviews, fostering more human interactions, and developing innovative platforms adapted to the AI age. Further research on responsible AI assistance in peer reviews is emphasised to maintain scientific integrity.
