Nearly 3,000 peer-reviewed medical papers have fake citations, AI-assisted audit finds
Key Points:
- A Columbia University School of Nursing AI-assisted audit found nearly 3,000 peer-reviewed biomedical papers contain fabricated citations that do not exist in scientific databases, revealing a significant problem in academic publishing.
- The study, published in The Lancet, analyzed 2.5 million papers from PubMed Central's Open Access between 2023 and early 2026 using an AI verification system, identifying 4,046 fake citations across 2,810 papers.
- The incidence of fake citations has increased more than 12-fold since 2023, with a sharp rise starting mid-2024, coinciding with the increased use of AI writing tools in academic research.
- Researchers warn that fake citations undermine clinical guidelines and patient care, as medical professionals may base treatment decisions on nonexistent evidence, and recommend publishers verify references and indexing services improve metadata for accuracy.
- The authors urge research integrity databases to track fake references systematically and call for retroactive screening and corrective actions on existing publications, noting that 98.4% of affected papers had not yet been addressed by publishers.