As artificial intelligence helps researchers produce academic papers faster, the pressure is shifting to another critical part of science: peer review.
Daniel Zantedeschi
A new study co-authored by Daniel Zantedeschi, assistant professor of information systems at the University of South Florida, proposes using carefully governed AI to help academic journals manage that growing workload — without replacing human reviewers.
Published in Management Science, the study points to signs that the strain is already increasing. Submissions to the journal rose 22% from 2024 to 2025, while generative AI is contributing to greater researcher productivity and a growing volume of scientific papers.
The researchers propose a journal-controlled AI system that would conduct an initial assessment of manuscripts. A journal-specific large language model could flag potential issues and allow authors to respond before the paper reaches human reviewers. Its use would be transparent, monitored, and subject to human oversight.
Using a mathematical model of journal operations, the researchers found that governed AI could, under certain conditions, increase review capacity, shorten turnaround times, and improve decision accuracy.
But the study also identifies an important risk: AI-generated errors could influence multiple reviewers in similar ways. That possibility makes governance and human oversight central to the approach rather than optional safeguards.
Although developed in the context of academic publishing, the model has broader applicability to industries that depend on expert review and face growing volumes of complex submissions.
Authors: Hemant K. Bhargava and Pantelis Loupos, UC Davis; Sarah H. Bana, Chapman University; Zhe Zhang, UC San Diego; Laura Brandimarte, University of Arizona; Vidyanand Choudhary, UC Irvine; J. Frank Li, University of British Columbia; Daniel Zantedeschi, University of South Florida.