The Scientist readers share their thoughts on the use of AI in peer review.
AI tools have become increasingly common in research, from multi-agent AI systems developed to automate the scientific process to tools that assist in writing. Now, questions about whether and how they should be used in peer review have emerged. We asked The Scientist’s readers to share their thoughts on AI’s role in evaluating scientific research. Here is what they had to say.
Thomas Dougherty, Consultant, Antibiosis Science Consulting, LLC.
I have been on the editorial board of one of the American Society for Microbiology journals for 21 years. Even with AI assistance, I have never employed AI to review a manuscript. There is too much work put into most studies by the authors to use what we all know is an imperfect set of tools to make recommendations. Further, some publications critical to evaluations are behind journal firewalls, and much of older literature is also not available to AI. This almost certainly means the AI tool never had access to that training information. Generative AI is an extremely complex statistical text generator. It can certainly develop and link some concepts. But it has no real-world access or experience beyond text and photos, which are broken down into tokens and used to adjust weights in the AI models. Authors deserve to have their work evaluated by other scientists with experience and history in the field.
Joseph Walker, Quality Assurance Specialist, NewAge Industries, Inc.
If you use AI as a tool, yes. However, if you use it as a standalone or final review, no.
Aram Megighian, Neuroscientist, University of Padova
Considering that it is also used for planning and writing papers, I have a personal and old idea that could be adapted to these times. In the past, research discoveries were discussed between scientists in the meetings of famous societies where they used to present their results. Why not return to those times? Imagine an online website where you can upload your manuscript without any editorial rule. A peer review analysis on the methods and results statistics and the paper is open for online discussion like a blog.
Nikolas Fokialakis, Pharmacognosist, University of Athens
I think AI is useful only as a helping tool.
Dana Marshall, Biologist, Meharry Medical College
I tend to lean towards using it. It has a much greater depth of knowledge through drawing from numerous resources. I wouldn’t have it as the only review, but I like the idea of using it as a reviewer along with at least one human.
Arijit Ghosh, Consultant, Fitness Fortitude
AI in peer review is a genuinely mixed bag, and I think the right take depends on where exactly in the process it is being used. AI is most defensible as a pre-check, flagging things for a human reviewer to look at rather than as a substitute for the reviewer’s judgment. The moment AI is writing the substantive critique instead of assisting one, the epistemic function of peer review starts to erode.
Responses have been edited for length and clarity.
Leave a Reply