Detecting AI-Generated Content in Academic Peer Reviews

Summary

Applies a detector trained on pre-LLM reviews to later review cycles at ICLR and Nature Communications, giving a clean temporal prevalence estimate: near-zero before 2022, then roughly 20% of ICLR reviews and 12% of Nature Communications reviews classified as AI-generated in 2025, with the sharpest NC jump between Q3 and Q4 2024. Extends the ICLR/EMNLP sentence-level estimates already in the KB (Liang et al.) into a journal setting and up to 2025.

Contribution

Queued — not yet read.

Method

Queued — not yet read.

Relevance to RISE

Surfaced by the 2026-08-29 literature sweep as a medium-priority candidate for the RISE knowledge base (angle: AI in peer review). Verified against: arXiv abstract page opened 2026-08-29 (title, 2 authors, v1 30 Jan 2026, full abstract with prevalence figures read)

Critique / open questions

Queued — not yet assessed.

Key quotes

The growing availability of large language models (LLMs) has raised questions about their role in academic peer review. (abstract)

This study examines the temporal emergence of AI-generated content in peer reviews by applying a detection model trained on historical reviews to later review cycles at International Conference on Learning Representations (ICLR) and Nature Communications (NC). (abstract)