This AI can improve your peer review — and make it more polite

Nature

by Nicola Jones

February 23, 2026

AI-Generated Deep Dive Summary

A new AI system designed by researchers at Stanford University aims to enhance peer review quality by making feedback more constructive and polite. The system, called the Review Feedback Agent, uses five large language models (LLMs) to analyze existing reviews and suggest improvements. This innovative approach addresses a common issue in academic publishing: peer reviews that are often vague, unprofessional, or even incorrect. By providing actionable advice, the AI helps reviewers offer clearer and more helpful feedback, potentially strengthening the quality of scientific research. The problem with current peer reviews is significant. According to James Zou, lead researcher on the project, nearly 13% of conference paper authors label reviews as poor due to their lack of specificity or professionalism. Reviews may include harsh comments like "these authors don't know what they're talking about" or overly simplistic critiques such as "not novel." These shortcomings can discourage researchers and undermine the peer review process, which is crucial for maintaining research quality. Zou's team addressed this challenge by curating a dataset of problematic reviews and appropriate responses. They trained their AI system to identify areas where feedback could be improved and provided suggestions accordingly. For instance, the AI might suggest replacing a vague comment like "not novel" with a more specific observation about how the paper could be strengthened. This approach was tested on 20,000 existing reviews ahead of the 2025 International Conference on Learning Representations in Singapore. The results were promising. In most cases, the AI suggested ways for reviewers to be more specific and constructive, often phrasing suggestions like "to make this feedback more actionable." The system avoids unhelpful or unprofessional language while ensuring reviews are detailed enough to guide authors effectively. This could lead to better revisions and stronger research papers. Ultimately, improving peer review quality benefits the scientific community as a whole. Peer reviews play a critical role in validating research and maintaining academic standards. By making this process more effective and respectful, AI tools like Zou's system could enhance collaboration and innovation in science. While it remains unclear whether these improvements translate into better research outcomes, the potential for more constructive feedback is a significant step forward in scientific communication. This development highlights how AI can support rather than replace human efforts in complex tasks

Verticals

scienceresearch

Originally published on Nature on 2/23/2026