More than 250,000 cancer research papers published over the past 25 years show writing patterns that match those of known fraudulent "paper mill" studies, according to a new analysis published in The BMJ. The finding exposes a massive integrity problem in the scientific literature that could mislead drug development and patient care.
How the AI detection system works
Researchers at Queensland University of Technology (QUT) and collaborators trained a language model called BERT to identify subtle textual fingerprints that appear in papers later retracted for fabrication. The system, an example of AI for Science & Research, analysed 2.6 million research papers indexed in cancer research between 1999 and 2024. When tested on verified examples, the model correctly flagged suspicious papers 91% of the time.
"We've essentially built a scientific spam filter," said Professor Adrian Barnett, from the School of Public Health and Social Work at QUT and the Australian Centre for Health Services and Innovation. "Just like your email system can spot unwanted messages, our tool flags papers that match the writing style and structure we see in retracted, fraudulent work."
The scale of suspicious cancer papers
The proportion of flagged papers rose sharply, from about 1% in the early 2000s to more than 16% in 2022. The problem appeared across thousands of journals, including those with strong reputations, and was especially concentrated in molecular cancer biology and early-stage laboratory research. Certain cancer types - gastric, liver, bone and lung - had particularly high rates of suspect studies.
"Paper mills are companies that sell fake or low-quality scientific studies. They are producing 'research' on an industrial scale, and our findings suggest the problem in cancer research is far larger than most people realized," Barnett said.
Journals begin testing the tool
Three scientific journals are already trialling the system as part of their editorial screening process, hoping to catch potential fakes before they reach peer reviewers. The researchers plan to adapt the tool for other fields and expect accuracy to improve as more confirmed paper mill cases are documented. They caution that flagged papers are warning signals, not proof of misconduct, and still require human expert review.
Why this matters for Science and Research professionals
Fabricated cancer research can infiltrate clinical guidelines and drug development pipelines, wasting time and resources. For researchers, the tool offers a way to triage the literature and avoid building on questionable data. For journal editors and peer reviewers, it provides a scalable first-pass filter to protect the integrity of the scientific record. As paper mills grow more sophisticated, automated detection systems like this one will become essential for maintaining trust in published research.
Your membership also unlocks:
