RightSignal
1 of 6

Your RightSignal

AI

Twice-Told Critiques: Claim-Level Homogenization of Peer Review in the LLM Era

Ilya Gusev · Submitted 19 Nov 2026 · 10 upvotes yesterday
Preprint · not peer reviewedTrending

The author measured how much the critiques in different reviews of the same paper overlap, across AI conferences and journals, and ran an AI-text detector over all reviews.

Key finding: At ICLR, overlap rose from 12.0% in 2026 to 21.3% in 2027, and a third review added 5.46 new problems instead of 6.12. Reviews flagged as AI-written overlapped more with each other.

Why it matters: Peer review works because different readers spot different problems. If reviews converge, more flaws may slip through.