AI in science – dusk or dawn?

Abstract

The peer review process remains a cornerstone of scientific integrity, ensuring that research findings are critically evaluated before entering the scientific record. With the growing integration of artificial intelligence (AI) and the widespread adoption of large language models (LLMs) such as ChatGPT, the research and publishing landscape is undergoing rapid transformation. While AI offers considerable advantages – enhancing efficiency in manuscript drafting, editing, and preliminary evaluation – it also introduces significant risks, particularly when used beyond its optimal scope. This viewpoint underscores the limitations of generative AI, including the phenomenon of “hallucinated” references and the inability to perform genuine critical thinking. These shortcomings raise serious concerns about the validity of scientific content when AI is used without appropriate human oversight. Emphasis is placed on preserving the human-centered nature of peer review, which is vital to safeguarding scientific credibility. In doing so, this article reinforces the necessity of evolving editorial and publishing policies, such as Elsevier’s updated guidelines on the use of generative AI, to ensure responsible integration of these technologies into the research ecosystem.

Keywords: artificial intelligence; ChatGPT; large language models; peer review