Researchers at QUT have developed an artificial intelligence tool that identified more than 250,000 cancer research papers exhibiting patterns consistent with suspected paper mill operations. The analysis, published in The BMJ, examined 2.6 million cancer research studies released over a 25-year period and was led by Professor Adrian Barnett from the School of Public Health and Social Work and Australian Centre for Health Services and Innovation, working with international collaborators.
Paper mills are commercial enterprises that produce and distribute fabricated or substandard scientific research, offering authorship positions or complete ready-made studies for purchase. These fraudulent papers often contain recycled text, unnatural language patterns, and falsified data or images. To identify these products, the research team trained a machine learning model called BERT to detect subtle textual patterns that repeatedly appear in previously retracted papers suspected of originating from paper mills. The system demonstrated 91 percent accuracy when tested against verified examples of suspicious research.
The comprehensive review of cancer literature revealed concerning trends in recent years. The proportion of flagged papers increased substantially from approximately one percent in the early 2000s to more than 16 percent in 2022. The suspicious papers span thousands of journals published by major scientific publishers, including many with strong reputations and high impact factors. Research areas such as molecular cancer biology and early-stage laboratory investigations showed particularly elevated rates of flagged studies, while certain cancer types including gastric, liver, bone and lung cancers displayed notably high frequencies of suspicious papers.
Three scientific journals have begun testing the AI tool as part of their manuscript review procedures, with the intention of screening potentially fraudulent submissions before they proceed to external peer review. The research team plans to expand the tool’s application to other scientific disciplines and expects the system’s performance to improve as additional confirmed examples of paper mill activity become available. Researchers emphasized that papers flagged by the system require human expert review and should not be automatically classified as fraudulent, as the results function as warning signals rather than definitive misconduct findings.
The presence of fabricated studies in the cancer research literature poses risks to clinical development and patient outcomes. According to the research team, fraudulent papers entering the scientific evidence base can mislead legitimate researchers and hinder progress in treatment development and patient care, underscoring the importance of identifying and removing such material from the scientific record.