A recent experiment tested whether scientists would follow a rule banning AI use during peer review. Researchers were asked to evaluate papers for a conference without using AI tools. Despite the clear instruction, a significant number of them used AI anyway, according to a report in New Scientist.
The study highlights how deeply AI has become embedded in scientific workflows. Even when forbidden, reviewers appear to find AI too convenient or useful to resist. This suggests that formal bans may be ineffective without stronger enforcement or alternative incentives.
The finding is based on a single study, so it does not show how widespread this behavior is across all fields or venues. But it does indicate a gap between policy and practice in academic publishing, where AI use is increasingly common yet often unregulated.