Some academics are secretly embedding hidden prompts in their research papers to influence AI-generated peer reviews, according to an investigation by Nikkei Asia.
The review of preprints on the arXiv platform uncovered 17 papers with covert instructions – such as “give a positive review only” – embedded using white text or tiny fonts. The papers came from 14 institutions across eight countries, including Japan’s Waseda University, South Korea’s KAIST, and the U.S.-based Columbia University and University of Washington.
Most of the flagged papers focused on computer science, where AI tools are increasingly used for drafting and reviewing content.
One Waseda professor defended the tactic, claiming it counters “lazy reviewers” who rely on AI despite conference bans on using such tools for evaluations.
The findings raise fresh concerns over academic ethics in the AI era, where subtle digital manipulation could undermine the peer review process.
