Updated
Updated · CBS New York · May 13
Audit Finds 4,000 Fake AI Citations in Biomedical Papers as Errors Jump 12-Fold
Updated
Updated · CBS New York · May 13

Audit Finds 4,000 Fake AI Citations in Biomedical Papers as Errors Jump 12-Fold

2 articles · Updated · CBS New York · May 13
  • More than 4,000 citations to non-existent studies were found across nearly 3,000 biomedical papers in an audit of millions of articles, according to a Lancet report led by Columbia nursing professor Maxim Topaz.
  • AI tools generated the bogus references when authors used them to polish manuscripts or supply supporting citations, sometimes inventing entire studies and other times attaching fake research to real authors.
  • None of the fabricated citations identified in the audit had been corrected or retracted, Topaz said, leaving them in the literature where they can feed into clinical guidelines and patient-care decisions.
  • The problem is accelerating: Topaz said fake references in published medical literature have risen 12-fold over the past three years and may represent only the "tip of the iceberg" across research fields.
With thousands of fake studies polluting medicine, is the scientific record now permanently compromised?
If a patient is harmed by AI's fake research, who is liable: the doctor, publisher, or the tech company?

Biomedical Research Faces Crisis as AI-Generated Fake Citations Increase 12x Since 2023

Overview

Biomedical research is facing a serious threat as AI-generated fake citations are rapidly increasing. This issue was uncovered by Columbia University researchers after Maxim Topaz, who often used AI writing tools, discovered a fake citation in his own submitted paper. Embarrassed by this, he and his team conducted a large audit, revealing that fabricated references are widespread and growing fast in scientific literature. Their findings, published in May 2026, show that the problem is not isolated but systemic, highlighting the urgent need for better detection and prevention to protect the integrity of research.

...