New Delhi: A recent study has found that nearly 3,000 peer-reviewed biomedical papers from across the world published since 2023 contain fake citations that do not correspond to actual papers.
For Maxim Topaz, associate professor at Columbia University’s School of Nursing and Data Science Institute, the discovery of a fabricated reference was the start of a large-scale audit. His team’s work was published as a peer-reviewed research letter in The Lancet on 7 May. Surprisingly, almost all the papers Topaz’s team flagged as carrying fake citations remain unchanged.
“98.4% of the papers we flagged remain in the literature without correction. Most people assume someone is checking, but actual action remains very rare,” Maxim Topaz told ThePrint.
AI at fault
The research team developed a verification system using AI, which scanned 2.5 million papers published from 1 January 2023 to 18 February 2026, in PubMed Central’s Open Access.
Among the 97.1 million references verified across PubMed, Crossref, OpenAlex and Google Scholar, the audit identified 4,046 fake citations across 2,810 research papers. In 2023, one in 2,828 papers contained at least one fabricated reference. By 2025, this rate had grown to nearly one in 458 research papers. In 2026, 1 in 277 papers had a fabricated reference.
“A more than 12-fold increase is not sloppiness. It is a systemic shift,” said Topaz. His team also noted that the spike in fabricated citations coincided with the rise of AI writing tools, which emerged in 2024.
Topaz agrees that in this case, AI is likely to be both the source of the problem and the solution.
“AI writing tools can generate fabricated references, and AI-powered verification tools can catch them. But the core of our detection is not AI judgment. It is a binary check against four databases covering more than 250 million scholarly works: Does this paper exist or not?” he said.
Also read: Hyderabad start-up TakeMe2Space turns to SpaceX after ISRO PSLV mission failure
Academia must catch up to AI
According to Topaz, it is unreasonable to expect reviewers to verify every reference. Instead, his team suggested that publishers integrate automated reference verification into submission workflows before peer review begins.
“The tools exist. The barrier is institutional, not technological,” said Topaz.
Topaz cautions that most research ethics training has not caught up with the reality that AI tools can introduce fabricated content into your work without you realising it.
“91% of affected papers had only one or two fabricated references. Many of these are likely honest mistakes by authors who used AI tools without verifying the output,” said Topaz.
More concerning is the fact that medical professionals make treatment decisions based on clinical guidelines. There is no way for them to know whether the study they are relying upon is backed by actual evidence.
“For example, one paper we reviewed had 18 out of 30 fake references. Some of those citations are already being cited by other papers and appear in systematic reviews that inform clinical care,” said Topaz.
While AI makes for a great assistant, Topaz highlights that it is not a replacement for one’s judgement.
“If we do not act, fabricated references will continue to propagate into systematic reviews and clinical guidelines, compromising the evidence chain that clinicians rely on,” he said.
(Edited by Prasanna Bachchhav)