No Hidden Prompts Needed! You Can Game AI Peer Review with Presentation-Only Revisions

In three linesStudy demonstrating AI reviewers can be manipulated through presentation-only changes (abstract, framing, narrative structure) without altering methods or results. Attack success rate of 75.1% with mean score gain of +1.21/10. AI reviewers confuse appearance with substance, favoring strength highlighting over weakness refutation.

Summary generated by Claude — human-verified