Overview
- The Columbia-led audit, published Thursday in The Lancet, scanned 2.5 million PubMed Central open-access papers and verified 97.1 million references to flag 4,046 fabricated citations across 2,810 papers.
- The share of papers with fake citations rose from about 1 in 2,828 in 2023 to 1 in 458 in 2025 and 1 in 277 in early 2026 after a sharp mid-2024 jump.
- The authors link the rise to growing use of AI writing tools that can hallucinate realistic-looking references that do not exist.
- Proposed fixes include automated checks before peer review, integrity flags in indexing records, retroactive screening, and retractions when fabrications support core claims, with about 98% of flagged papers still lacking publisher action at the audit date.
- Reactions split on remedies as JAMA veterans urged retraction in any case, other experts backed case-by-case corrections and clearer methods, and publishers such as Taylor & Francis and PLOS described or explored system-wide screening.