Celebrating Peer Review Week 2025
Peer Review Week is 15–19 September — an opportunity to thank the dedicated volunteers so critical to scientific publishing, and to explore key issues in the field. Visit our Peer Review Week page to help celebrate all our reviewers and see other blog posts on this topic!
As AI tools become more sophisticated and accessible, their potential to reshape the peer review process grows, and so do the questions about how, when, and where they should be used. In this second post in our Peer Review Week series, we ask: Are there ways AI can be incorporated into the peer review process while still keeping the process centered on accuracy and scientific progress? And just as importantly, what aspects of peer review must remain uniquely human?
“AI may be an assistant, but it cannot be a peer. Peer review, at its core, is about accountability, mentorship, and the human voice in science, things that remain irreplaceable.”
— Shubhechchha Sharma
The reviewers featured here are recipients of the AMS Editor’s Award. They share thoughtful insights reflecting both optimism and caution, grounded in a shared commitment to scientific rigor. As technology evolves, so too must our approaches — with care, transparency, and clarity about the irreplaceable role of human expertise.
Rosimar Rios-Berrios, NSF National Center for Atmospheric Research (NCAR):
“Evaluation of the quality, innovation, and uniqueness of scientific studies must remain uniquely human because these qualities vary from study to study and cannot be easily generalized. Perhaps AI tools can be incorporated for screening studies for plagiarism and for improving the written quality of the scientific text to make it accessible to wide audiences.”
Matthew Brown, National Severe Storms Laboratory:
“I do think there is a place for AI within the review process. AI can be used to explore foundational literature on less familiar topics, quickly summarize results in confusing or complex papers, bridge language barriers for non-English authors and reviewers, and help neurodivergent reviewers synthesize their feedback, among other things. The crux for me, as a scientist and a peer reviewer, is, are you using AI as a tool to facilitate deeper understanding or a means to bypass the critical thinking process? If the answer is the latter, then you are not providing any unique insight, or at least any insight that the authors couldn’t get on their own by feeding the paper through an AI program to critique their manuscript. The active dialogue between authors and reviewers is crucial to maintaining high-quality science.”
Scott Landolt, NSF National Center for Atmospheric Research (NCAR):
“AI can be used in the review process to help search for additional citations that may be applicable to a manuscript, or to help reword sections for clarity. Ensuring that the scientific methods, data acquisition and quality control, and resulting conclusions are robust and repeatable are something that should remain uniquely human to the review process.”
Shubhechchha Sharma, University of Minnesota:
“The heart of peer review must remain human. Reviewing originality, theoretical contributions, or paradigm-shifting ideas requires creativity and deep disciplinary expertise, qualities no algorithm can replicate. The same goes for weighing ethical implications, ensuring research integrity, and safeguarding fairness. These responsibilities demand lived experience, cultural understanding, and the ability to see beyond the text into its broader meaning. Most importantly, only human reviewers can make final publication decisions and provide developmental feedback that not only improves individual papers but also nurtures authors and strengthens the scientific community as a whole. AI may be an assistant, but it cannot be a peer. Peer review, at its core, is about accountability, mentorship, and the human voice in science, things that remain irreplaceable.”
AMS has adopted guidelines to inform the use of AI tools for authors (see #12) and reviewers (see #13). The AMS Publications Commission welcomes your thoughts on the policies. Email [email protected].
