AI Is Transforming Peer Review — and Many Scientists Are Worried

Summary

A Nature feature surveying how AI — particularly LLMs — is already being used in peer review. The article opens with ecologist Timothée Poisot identifying a referee report that was likely generated by an LLM (it contained the giveaway line "Here is a revised version of your review with improved clarity and structure"). It then catalogues the range of current deployments: publishers using AI to check statistics, summarise findings and select reviewers; new sites offering one-click AI reviews; and a Wiley survey of ~5,000 researchers in which 19% said they had already used LLMs to "increase the speed and ease" of their reviews. A separate study of AI-conference reviews from 2023–2024 estimates 7–17% of reports were "substantially modified" by LLMs.

Contribution

A widely-cited journalistic synthesis of the early-2025 state of play on AI in peer review, capturing the policy patchwork (many funders and publishers forbid reviewer LLM use citing confidentiality leakage, while some quietly encourage it) and the normative split between enthusiasts and critics. Acts as the news-side companion to the Nature editorial on transparent peer review.

Method

Feature journalism; combines interviews (Poisot, Bergstrom, Porsdam Mann, Gruda) with citation of the Wiley researcher survey and a study of AI-conference reviews.

Relevance to RISE

Provides current journalistic context for the AI-peer-review pillar of RISE — the policy environment, the existing reviewer-side behaviours, and the ethical stakes that any RISE reviewer project (notably reviewer) is entering. Useful citation for any RISE writeup that needs to motivate the problem to a non-specialist audience.

Critique / open questions

A news feature, not a peer-reviewed empirical study; the underlying numbers are second-hand (a Wiley survey, a single AI-conference study). Selection of voices is editorial.

Key quotes

"It contained the telltale sentence, 'Here is a revised version of your review with improved clarity and structure', a strong indication that the text was generated by large language models."

"In a survey of nearly 5,000 researchers, some 19% said they had already tried using LLMs to 'increase the speed and ease' of their review."