Who reviewed this? Toward responsible integration of large language models for peer review of scientific articles in dental medicine

Abstract

The introduction and advancement of large language models (LLMs), such as ChatGPT, DeepSeek, and Google Gemini, present both opportunities and challenges for peer review in dental research. In this article, we propose a framework to inform the discourse on the responsible use of LLMs in dental peer review. We conducted a cross-sectional review of peer review policies from the top 50 dental journals, based on their 2024 Journal Impact Factor, to assess current guidance on LLM use. Our analysis revealed variability across dental journals: some journals permit restricted LLM use under specific conditions, while many either prohibit their use or lack explicit policies. Key concerns regarding LLM use identified by the authors include potential breaches of confidentiality, ambiguity in authorship, reduced reviewer accountability, and inherent limitations of LLMs in terms of domainspecific expertise and factual accuracy. Our proposed framework addresses confidentiality safeguards, suggested appropriate LLM applications, areas requiring caution, disclosure requirements, and accountability standards. It emphasizes that reviewers retain full responsibility for all submitted content, irrespective of LLM assistance. To protect confidentiality, the framework encourages offline or locally hosted LLMs. It also recommends regular policy reviews and reviewer training. This framework aims to support the thoughtful adoption of LLMs in dental research publishing. When employed judiciously, LLMs offer potential benefits in improving review clarity and efficiency, particularly for reviewers writing in a non-native language. However, their use must be grounded in clear ethical principles to ensure the integrity of dental peer review.