Investigators Find Rising Rate of Fabricated Citations in Biomedical Literature
KEY TAKEAWAYS
- An audit of more than 2.4 million biomedical papers identified 4046 fabricated references across 2810 publications.
- The rate of papers with at least 1 fabricated citation increased sharply from 2023 to early 2026.
- The authors recommend automated reference verification before peer review to protect the integrity of evidence synthesis.
Results from an audit of reference integrity published in The Lancet found that fabricated references are present in peer-reviewed biomedical literature and appear to be increasing over time. The analysis highlights a growing research-integrity challenge in which references that appear plausible—often with real author names, credible dates, and correct formatting—point to nonexistent publications. For neurology, where clinical guidelines and practice updates frequently rely on systematic reviews and rapidly expanding evidence bases, routine reference verification may be especially relevant to preserving confidence in cited evidence.
The authors developed an automated reference-verification system to evaluate PubMed Central’s Open Access subset from January 1, 2023, through February 18, 2026. The audit included 2,471,758 biomedical papers and 125,615,773 structured references. Of these, 97.1 million references carried a PubMed identifier and were verified against PubMed, Crossref, OpenAlex, and Google Scholar. References that passed automated mismatch filters and could not be found in any database were classified as fabricated; those linked to incorrect identifiers but matching real publications were classified as reference errors.
Research Integrity Gaps Revealed by the Audit
- Investigators identified 4046 fabricated references across 2810 papers.
- In 2023, approximately 1 in 2828 papers contained at least 1 fabricated reference.
- By 2025, approximately 1 in 458 papers contained a fabricated reference, and during the first 7 weeks of 2026, 1 in 277.
- The fabrication rate rose from approximately 4 per 10,000 papers in 2023 to 51.3 per 10,000 papers in the fourth quarter of 2025 and 56.9 per 10,000 papers in early 2026.
- Review articles had a 57% higher fabrication rate than other publication types.
- Most affected papers (91%) contained 1 or 2 fabricated references, but 246 papers contained 3 or more.
The authors noted that the analysis could not determine cause. Potential causes they cited include paper mill activity, intentional misconduct, and uncritical use of artificial intelligence tools. Limitations in the audit include the exclusion of references without PubMed identifiers, use of the PubMed Central Open Access subset rather than the full biomedical literature, and a validation design that estimated precision but not recall.
The commentary recommends integrating automated reference verification into journal submission workflows, adding integrity metadata to article records, retroactively screening existing publications, and establishing fabricated references as a discrete category in research-integrity databases.
Source
Topaz M, Roguin N, Gupta P, et al. Fabricated citations: an audit across 2.5 million biomedical papers. The Lancet. 2026;407:1779-1781.