Fake Citations Surge in Biomedical Research Since 2023

By VirentaNews Staff — May 10, 2026

An alarming new audit of 2.5 million biomedical research papers has uncovered a steep rise in fabricated citations, with one in every 500 references now estimated to be entirely fictitious. The study, which analyzed 97 million citations published between 2010 and 2025, found that the rate of fake citations has increased more than fourfold since 2023, from 0.15% to 0.22% of all cited references—a seemingly small figure that translates to hundreds of thousands of misleading citations across the scientific literature. These phantom references, often to non-existent journals or papers with plausible but fabricated titles and authors, are not random errors but deliberate fabrications designed to bolster the credibility of new research without contributing actual evidence. The findings, published in Nature, suggest that the integrity of the scientific record is under growing pressure from both human misconduct and the unchecked use of AI-generated text in scholarly writing.

The Growing Crisis in Scientific Trust

The credibility of scientific research rests on a foundation of verifiable evidence, with citations serving as the connective tissue linking new discoveries to prior knowledge. But as the volume of published research has exploded—over one million new biomedical papers are published annually—the system of peer review and editorial oversight has struggled to keep pace. The new audit, conducted by a team at the University of Edinburgh and the Allen Institute for AI, reveals that the rate of citation fraud has accelerated sharply since 2023, coinciding with the widespread adoption of generative artificial intelligence tools in academic writing. These tools, while capable of summarizing literature and drafting text, often ‘hallucinate’ references—generating plausible-sounding citations that do not exist. Researchers under pressure to publish may then include these fake citations without verification. The result is a self-reinforcing cycle in which false references are cited by subsequent papers, embedding misinformation deeper into the scientific record.

How the Audit Uncovered Systemic Fraud

The investigation began as a routine effort to map citation networks in biomedicine but quickly revealed an anomaly: thousands of papers referenced studies that could not be located in any major database, including PubMed, Scopus, or CrossRef. Using a combination of natural language processing and manual verification, the team flagged citations with suspicious patterns—such as journals with names nearly identical to legitimate publications, or authors with names that appear only once in the literature. Over 180,000 such citations were identified across the 2.5 million papers analyzed, with the most dramatic increase occurring in low-impact journals and papers originating from institutions with limited oversight. Notably, many of the fake citations were clustered in papers discussing niche topics like rare genetic disorders or novel drug mechanisms—areas where fewer experts exist to spot inconsistencies. The researchers also found that a significant portion of the fabricated references appeared in papers that used AI-assisted writing tools, suggesting a direct link between generative AI and citation pollution.

Why Fabricated Citations Undermine Science

The inclusion of fake citations is more than an ethical lapse—it distorts the scientific process by creating false narratives of consensus and evidence. When a fabricated reference supports a claim about drug efficacy or disease mechanism, it can mislead clinicians, policymakers, and future researchers. The audit found that some fake citations had been cited multiple times, creating the illusion of established knowledge. For example, one non-existent paper on CRISPR-based gene therapy for cystic fibrosis was cited in at least 17 subsequent studies, none of which questioned its validity. Experts warn that such ‘citation cascades’ can persist for years, especially in fast-moving fields where researchers rely heavily on bibliographies rather than primary sources. The problem is compounded by the lack of automated tools to detect citation fraud: while plagiarism detection is routine, no widely used system verifies the existence of every cited paper. As a result, the scientific community may be building on a foundation partially constructed from fiction.

Who Is Most at Risk?

The ripple effects of citation fraud are felt most acutely in clinical medicine, drug development, and public health policy, where decisions are often based on systematic reviews and meta-analyses that aggregate findings from hundreds of studies. If even a small fraction of the cited literature is fabricated, the conclusions of these high-impact reviews may be compromised. Early-career researchers, particularly those in countries with less robust research infrastructure, are also vulnerable—both as unintentional perpetrators who rely on AI tools and as targets of predatory publishing practices that incentivize citation manipulation. Journals with weak editorial standards are disproportionately represented in the dataset, suggesting that the problem is concentrated in parts of the publishing ecosystem that lack rigorous screening. However, even high-impact journals are not immune: the audit identified several fake citations in papers published in well-known titles, underscoring the need for universal verification protocols.

Expert Perspectives

“This isn’t just about bad actors—it’s a systemic failure,” says Dr. Elena Martinez, a research integrity specialist at the University of Oxford. “The pressure to publish, combined with the ease of AI-generated text, has created a perfect storm.” Others argue that the focus should shift from punishment to prevention. Dr. Kwame Nsiah, a bioethicist at McGill University, emphasizes that “many researchers aren’t deliberately cheating—they’re overwhelmed and under-resourced.” Some experts advocate for mandatory citation verification software, while others call for journals to require raw reference data in machine-readable formats to facilitate audits. There is growing consensus that both human oversight and technological solutions are needed to restore trust.

Going forward, the scientific community faces urgent questions: How can citation integrity be enforced at scale? Will AI tools be adapted to verify—rather than generate—references? And can publishers, institutions, and funders align on global standards for research transparency? The answers will shape the credibility of science in the decades to come.

Source: Nature