AI developed by University of Colorado Spots Fraudulently Published Journals

Researchers from the University of Colorado Boulder have developed an AI tool. It is designed to detect predatory scientific journals. These journals often charge authors high fees to publish research without conducting proper peer review.
The AI system analysed the websites of 15,200 open-access journals and flagged over 1,400 as potentially fraudulent. Human reviewers later confirmed that more than 1,000 of those titles showed signs of being predatory. The study was published in Science Advances on August 27, 2025.

AI focused on Key Indicators

The AI tool was trained using data from the Directory of Open Access Journals (DOAJ). This is a nonprofit organisation that has been assessing journal legitimacy since 2003. DOAJ uses six core criteria to evaluate journals, including peer review transparency, clear editorial policies, and credible editorial boards.
As reported by Science Daily, the AI system was developed to analyse publicly available data from journal websites. It looked for several indicators commonly found in predatory journals. These included fake or missing editorial board information and frequent grammatical errors. It also checked for high levels of self-citation and irregular publishing patterns.
According to Science Daily, the AI considered the number of authors’ institutional affiliations. It also examined the volume of articles published as potential warning signs. Journals that published an unusually high number of articles or featured authors with excessive affiliations were flagged for further review.

Also read: The New Frontlines: How AI is Redefining Warfare and Peace

AI findings were Validated by Researchers

After the AI flagged the initial 1,400 journals, human experts reviewed a subset of the data. They aimed to evaluate the system’s accuracy. According to Science Daily, the experts determined that around 350 journals were mistakenly flagged. This means the AI had falsely identified them as fraudulent. However, over 1,000 journals were confirmed to exhibit patterns consistent with predatory publishing.

The lead author of the study is Associate Professor Daniel Acuña. He is from the University of Colorado Boulder’s Department of Computer Science. Acuña stated that the tool is not intended to replace human judgment. Instead, it is designed to assist in prescreening large datasets. As quoted by Science Daily, Acuña said, “I think this should be used as a helper. It should help to prescreen large numbers of journals. But human professionals should do the final analysis.”

Nexus has Legs in many Countries

Acuña also noted that he regularly receives spam emails from unknown journals offering to publish his work for a fee. As reported by Science Daily, Acuña explained, “They will say, ‘If you pay $500. If you pay $1,000, we will review your paper.’ In reality, they don’t provide any service. They just take the PDF and post it on their website.”

The study found that predatory journals often target researchers in countries such as China, India, and Iran. Research institutions in these countries may be newer. Additionally, the pressure to publish is high.

As reported by Science Daily, the AI system is currently not publicly available. However, the research team plans to make it accessible to universities and publishers in the future. The researchers aimed to build a transparent and interpretable system, avoiding the “black box” issue found in some AI tools.

The study had several co-authors. One of them was Han Zhuang from the Eastern Institute of Technology in China. Another was Lizheng Liang from Syracuse University in the US.

For all latest updates follow theviralmail on X, Facebook, and Instagram