WASHINGTON (dpa-AFX) - A new review published in The Lancet found that thousands of references used in biomedical research papers may be fake.
'When those fake references are making it into the literature, they will end up in those guidelines, and that's how doctors decide how to provide care for you,' lead author Maxim Topaz, an associate professor at the Columbia School of Nursing, told CBS News. 'Your doctor could be making decisions around treatment based on studies that never existed.'
Researchers from Columbia University used artificial intelligence (AI) to examine 2.5 million research papers published between 2023 and early 2026. They checked 97.1 million references to see whether the cited studies actually existed and matched the information listed in the papers.
The AI system compared references with trusted databases such as PubMed, Crossref, OpenAlex, and Google Scholar. If a reference could not be found anywhere, it was labeled as fake. If the reference existed but had the wrong details or identifier, it was marked as an error. The system was found to be 91% accurate.
The study identified 4,046 likely fake references across 2,810 papers. The problem appears to be growing quickly. In 2023, about one in every 2,828 papers contained at least one fake reference. By 2025, that number rose to one in every 458 papers, and in the first weeks of 2026, it climbed further to one in every 277 papers.
Researchers also found signs of 'paper mills', businesses that produce and sell low-quality or fraudulent research papers. Some papers contained fake references related to topics such as CRISPR gene editing, AI-driven vaccines, and gut microbiome research. Most affected papers had only one or two fake references, but some had several. Review articles were more likely to contain fabricated references than other types of papers.
The researchers said the rise in fake references may be linked to the growing use of large language models (LLMs) and increased activity from paper mills. They also noted that the fake references looked convincing, with proper formatting, realistic dates, and real researcher names, making them difficult to spot.
To address the issue, the team recommended that publishers use automated reference-checking tools before papers are reviewed or published. They also suggested that journals should recheck older papers and correct or retract studies where fake references affect the findings.
Copyright(c) 2026 RTTNews.com. All Rights Reserved
Copyright RTT News/dpa-AFX
© 2026 AFX News
