Overview
- A PNAS paper led by Columbia University School of Nursing, published Tuesday, found fabricated or unverifiable references in physics, materials science, computer science, engineering and chemistry.
- The team scanned 500,000 papers (January 2023–March 2026) and ran 21.6 million DOI-bearing references through an automated checker that flagged 1,135 suspect citations across 619 papers.
- Two independent experts reviewed a random sample of 400 flagged items and confirmed about 85.8% were fabricated, and the measured rate rose from roughly 1 in 1,600 papers in 2023 to about 1 in 435 in early 2026.
- Field rates ranged from about 25.6 fake citations per 10,000 papers in physics to 4.3 per 10,000 in chemistry, which the authors scaled to estimate roughly 8,500 papers may contain at least one fabricated reference and noted only two flagged papers had been retracted.
- The researchers urge journals and indexers to adopt automated reference checks at submission, run retroactive screens of the literature, record fabricated references as a distinct integrity category, and are expanding audits — including screening millions of medical papers with HHS support — to guide editorial protocols.