Researchers have developed an artificial intelligence (AI) tool capable of identifying more than 1,000 potentially questionable open-access journals, highlighting a growing concern over the integrity of scholarly publishing. The study, published in Science Advances on 27 August, analyzed around 15,000 open-access journal titles for indicators of dubious publishing practices, including insufficient peer review and opaque editorial standards.
The AI tool flagged journals exhibiting red flags such as unusually fast article turnaround times, excessive self-citation, and editorial board members with weak institutional affiliations. Notably, none of the journals identified had previously appeared on any watchlists, and some were published by reputable companies. Collectively, these journals have produced hundreds of thousands of papers that have been cited millions of times, underscoring the scale of the issue.
“There’s a whole group of problematic journals in plain sight that are functioning as supposedly respected journals that really don’t deserve that qualification,” said Jennifer Byrne, a cancer researcher at the University of Sydney and co-author of the study.
The AI, currently available in a closed beta, is intended to assist publishers and indexing organizations in reviewing journal portfolios. However, the developers caution that the tool is not infallible and cannot replace detailed expert evaluations. “A human expert should be part of the vetting process before any action is taken,” said Daniel Acuña, a computer scientist at the University of Colorado Boulder.
The tool’s training relied on data from 12,869 legitimate journals indexed in the Directory of Open Access Journals (DOAJ), as well as 2,536 journals previously flagged for quality concerns. When applied to 15,191 journals in the Unpaywall database, the AI identified 1,437 as questionable, though some 345 were likely false positives. Researchers also noted that a stricter AI setting reduced false alarms but missed thousands of problematic journals, while a looser setting flagged nearly 8,800 titles but produced a high rate of false positives.
Cenyu Shen, deputy head of editorial quality at DOAJ, emphasized the importance of human oversight in the process. “AI could certainly play a useful supporting role, helping us manage scale and reduce the labour-intensive nature of reviews,” Shen said, while warning that automated tools may disadvantage non-English-language journals or editors from less-funded institutions.
The study highlights the challenges posed by the rise of “questionable open-access journals” and suggests that AI tools could help tackle the growing workload of journal assessments—provided they are used alongside rigorous human evaluation.
