In May 2026, The Lancet published an audit by Maxim Topaz and colleagues (Columbia University) that scanned 2.5 million biomedical papers and verified 97.1 million references with an AI-assisted system. They found 4,046 fabricated citations in 2,810 papers — references to studies that simply do not exist. The rate rose 12-fold since 2023: 1 in 2,828 papers in 2023, 1 in 458 in 2025, 1 in 277 in early 2026 — with the sharpest jump from mid-2024, coinciding with the spread of AI writing tools. One paper had 18 of 30 references fake. Review articles — the ones clinical guidelines are built on — run a 57% higher fabrication rate. Over 98% of affected articles had seen no publisher action. The authors couldn't prove cause — paper mills, misconduct, and uncritical AI use are all on the table — but the fake citations are already being cited by other papers and appearing in systematic reviews that inform clinical care. This 60-second check is the verification routine the study says everyone should be running. Free, no signup.
AI paper fraud is cybersecurity news because fake research poisons the well everyone drinks from — including security guidance. With 1 in 277 medical papers citing a study that does not exist, fabricated citations are being industrialized. This 60-second check teaches you to spot the red flags before you cite or act on a fake study.
This page checked one integrity failure. DisclosureGuard scans content for AI-generation signals and disclosure obligations — the automated verification layer the Lancet authors say publishing needs.
AI paper fraud is the use of generative AI to fabricate research papers, citations, or peer reviews at industrial scale. A 2026 analysis found 1 in 277 medical papers cites a study that does not exist. The red flags — tortured phrases, impossible citations, paper-mill authorship patterns — are learnable, which is what the check above teaches.
Topaz et al. (The Lancet, 2026) audited 2.5 million biomedical papers in PubMed Central's Open Access subset, verifying 97.1 million references with an AI-assisted system.
Not necessarily. The authors could not determine cause — possibilities include paper-mill activity, intentional misconduct, and uncritical use of AI tools.
They recommended publishers verify references with every submission using automated systems, indexing services add integrity metadata to article records, existing publications be screened retroactively, and research-integrity databases create a dedicated category for fabricated references to enable systematic tracking.