More than half of researchers now use AI for peer review, often against guidance (www.nature.com)

A recent survey reveals that over 50% of researchers now utilize AI in peer review processes, often contrary to established guidelines that discourage the uploading of unpublished manuscripts to third-party tools. While some publishers, like Frontiers, permit limited AI usage with required disclosures, there are significant concerns about confidentiality and the potential compromise of authors' intellectual property. The survey underscores the urgent need for publishers to adapt policies to address this growing trend and ensure responsible AI use in scientific publishing. The survey findings show that 59% of AI users in peer review employ it to craft reports, while 29% use it to summarize manuscripts and identify gaps, and 28% utilize it for misconduct flagging, including plagiarism checks. Experiments by researchers, such as Mim Rahimi at the University of Houston, indicate that while AI models like GPT-5 can replicate the format of peer review reports, they struggle with providing constructive feedback and may introduce factual errors. These developments provoke critical discussions on the efficacy and ethical implications of AI in the peer review landscape, highlighting the necessity for clear guidelines and human oversight in its application.