Home / Technology / AI Peer Review: Ban Flops as Scientists Cheat
AI Peer Review: Ban Flops as Scientists Cheat
28 Sep
Summary
- Most reviewers ignored AI use ban in scientific paper assessments.
- AI use had minimal impact on paper acceptance rates or scores.
- Reviewers cited workload and unclear rules for AI use.
At the 2026 International Conference on Machine Learning in Seoul, South Korea, a study explored the impact of banning AI in scientific peer review. Researchers found that restricting AI use for assessing papers made little difference to reviewer conclusions.
This was largely because many reviewers opted to use AI tools regardless of the prohibition. The experiment involved 24,661 submitted papers and 17,886 reviewers, comparing a strict AI ban with a permissive policy. Acceptance rates were nearly identical, with 27% under the strict policy and 26.5% under the permissive one.
An anonymous survey revealed that 22.5% of reviewers instructed not to use AI admitted to using it anyway. Reasons cited included large workloads and insufficiently clear rules. Researchers noted that AI text detectors also indicated a substantial usage under both policies, though they cautioned about detector accuracy.
The findings suggest that enforcing an AI ban in peer review is challenging due to academic workloads and the temptation to leverage AI for efficiency. This has implications for other conferences and fields facing similar issues.
