A sweeping audit of nearly 2.5 million biomedical papers has uncovered a sharp rise in made-up references within peer-reviewed research. Since 2023, the rate of fabricated citations has increased more than twelvefold, according to a study published in The Lancet by researchers at Columbia University and other institutions.
The team, led by Maxim Topaz, examined 2.47 million papers from the open PubMed Central archive, published between January 2023 and February 2026. They checked 97.1 million references in total. Of those, 4,046 references were flagged as fabricated, spread across 2,810 papers. A reference was considered fabricated if its listed title could not be found in any of four major databases: PubMed, Crossref, OpenAlex, and Google Scholar.
A sharp spike starting in mid-2024
The timeline shows a clear pattern. Throughout 2023, the rate remained steady at about four fabricated references per 10,000 papers. Starting in mid-2024, it began climbing quickly. By the end of 2025, it reached 51.3 per 10,000 papers, and in the first seven weeks of 2026, it hit 56.9 per 10,000. That represents more than twelve times the baseline.
The authors suspect a likely cause: the widespread adoption of language models like ChatGPT, which took off in late 2022. Because papers typically take 100 to 200 days from submission to publication, AI-generated text would not appear in PubMed Central in large numbers until mid-2024. The researchers did not rule out other possible contributors, such as increased paper-mill activity or changes in indexing practices.
Hard to spot and risky for clinical guidelines
The fake references are particularly troublesome because they are difficult to detect. They match the paper's topic, follow correct formatting, credit real researchers, and carry plausible publication years. In one urology paper, 18 of 30 checked references were fabricated, yet all closely matched the narrow surgical subject.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
The researchers also found patterns suggesting coordinated paper-mill activity. Two authors appeared in eleven papers from the same surgical journal, with a total of 15 fabricated references on topics such as CRISPR diagnostics and the gut microbiome.
At the time of the audit, 98.4 percent of the affected papers had received no response from their publishers. Review articles were hit hardest, showing a 57 percent higher fabrication rate than other paper types. That is especially concerning, the authors noted, because reviews often serve as the basis for clinical guidelines. If a guideline cites a paper with partly fabricated sources, the entire evidence chain behind treatment decisions can be compromised.
Scientific infrastructure needs to adapt
The scientific community has begun to respond, but the efforts remain uneven. Arxiv has tightened its sanctions for unchecked LLM output in manuscripts, including hallucinated sources, threatening offending authors with a one-year ban. An analysis of accepted NeurIPS 2025 papers had already shown that even top AI conferences cannot reliably catch fabricated citations. One possible countermeasure is CiteAudit, an open-source system for automated citation checking, although it also demonstrates how poorly commercial language models perform at identifying their own reference problems.
The researchers recommend four steps: automated reference checks before peer review, integrity metadata in article datasets, retroactive screening of already-published papers, and a dedicated "fabricated references" category in research integrity databases. The authors themselves used Claude for code development and grammar checking during the study.
Related on Neura Market
- AI Tools Directory: Explore tools for research integrity and automated citation checking.
- Automation Marketplace: Find workflows to streamline peer review and fact-checking.
- Research Platform Directory: Browse platforms for scientific publishing and data verification.