eLife journal, researchers, journal editors deepfake (Dec 2025)

A research team in China used Claude 2.0 to generate peer-review reports for 20 cancer-biology papers from eLife. The AI-generated reports fooled AI detection tools. The study found that Claude could write plausible citation requests and convincing rejection recommendations, raising the risk of journals rejecting good papers. Hydrologist at the Warsaw University of Life Sciences suspects that some referee reports were written by AI. Some journal editors are accepting AI-generated referee reports, unwittingly or otherwise, due to a shortage of peer reviewers.