Fake References Surge in Medical Papers, Study Finds
Researchers at Columbia University identified 4,046 fabricated citations in peer-reviewed medical papers, revealing a system vulnerable to fraud that could steer clinicians toward ineffective treatments. The study, published in The Lancet on May 9, examined 125.6 million references across 2.5 million papers posted to PubMed Central between January 2023 and February 2026.
The fake references appeared in 2,810 papers. Most contained one or two false citations, but 246 papers included three or more, suggesting deliberate manipulation rather than isolated errors.
Sharp Increase Tied to AI Writing Tools
The rate of papers containing fake references climbed from 4 per 10,000 throughout 2023 to 57 per 10,000 by early 2026. The sharpest spike began in mid-2024, coinciding with widespread adoption of generative AI writing tools.
Researchers used an AI verification system to flag references with digital object identifiers or PubMed database IDs that did not match actual published work. Kathryn Weber-Boer, a researcher at Digital Science, told Nature that while the source of fabricated citations remains unclear, "the recent upward trend suggests the influence of generative AI."
Direct Risk to Patient Care
When fake citations underpin a paper's scientific claims, the entire foundation weakens. One reviewed paper contained 18 false references out of 30 total citations, and some had already been cited in other papers and clinical guidelines.
Maxim Topaz, who led the Columbia team, said clinicians rely on these papers to make treatment decisions. "Papers that cite fake studies can steer patients toward the wrong treatments," he said.
Publishers took action on only 1.6% of the problematic papers identified during the investigation. In commentary accompanying the study, Howard Bauchner of Boston University School of Medicine and Frederick Rivara of the University of Washington School of Medicine called for stronger research ethics oversight.
Recommended Actions
The researchers proposed two steps:
- Publishers should verify references before accepting papers for publication
- Indexing services should provide additional information to help users assess reference accuracy
The findings underscore a structural weakness in academic publishing. As generative AI tools become standard writing aids, the ability to distinguish fabricated sources from legitimate ones has become essential for maintaining research integrity.
Your membership also unlocks:
