AI Hallucinations and Fake Citations Threaten Trust in Biomedical Research: Experts Warn

Written By :  Medha Baranwal
Medically Reviewed By :  Dr. Kamal Kant Kohli
Published On 2026-05-25 04:30 GMT   |   Update On 2026-05-25 05:43 GMT

USA: The integrity of scientific literature relies fundamentally on the accuracy of its references. Each citation is expected to direct readers to verifiable evidence supporting the claims made. However, two recent large-scale studies—“Fabricated citations: an audit across 2·5 million biomedical papers” by Maxim Topaz and colleagues, and “LLM hallucinations in the wild: Large-scale evidence from non-existent citations” by Zhenyue Zhao and team—have revealed a growing and concerning disruption to this foundation.

Published in The Lancet, the study led by Topaz represents one of the most comprehensive audits of reference integrity to date. The researchers analysed approximately 2.5 million biomedical papers and over 125 million references, focusing on those linked to identifiable records in major databases. Using an automated verification system, they compared citation details against databases such as PubMed, Crossref, OpenAlex, and Google Scholar to identify mismatches and non-existent entries.
Their findings showed that more than 4,000 fabricated references were embedded across 2,810 papers. Although this represents a small fraction of the total literature, the rate of occurrence has increased sharply. In 2023, only about one in 2,800 papers contained fabricated references; by 2025, this rose to approximately one in 458 papers, and early 2026 estimates suggest the problem is continuing to escalate.
Most affected papers contained one or two fabricated citations, but a smaller subset included multiple such references, sometimes forming a significant portion of the bibliography. Review articles were particularly vulnerable, showing a notably higher rate of fabricated references compared to other study types. The study also identified patterns that may indicate coordinated misconduct, including recurring author groups and repeated use of fabricated citations across multiple publications.
Complementing these findings, Zhao and colleagues examined over 111 million references across platforms including arXiv, bioRxiv, SSRN, and PubMed Central. Their study focused on “hallucinated” citations generated by large language models (LLMs), which are increasingly being used in academic writing.

Researchers estimated that nearly 147,000 non-existent citations were generated in 2025 alone, rising sharply with increased AI use. These hallucinated references often appear credible, with realistic titles, real author names, and proper formatting, making detection difficult.

Such errors were more common in AI-intensive fields and AI-assisted manuscripts, with smaller teams and early-career researchers disproportionately affected. Notably, these citations frequently credited well-established and male researchers, raising concerns about reinforcing existing academic biases.

Key Findings
  • Over 4,000 fabricated references identified across 2,810 biomedical papers
  • More than a 12-fold increase in fabricated citations from 2023 to 2025
  • By early 2026, nearly 1 in 277 papers contained at least one fabricated reference
  • The majority of affected papers had 1–2 fabricated citations, though some had several
  • Review articles showed a 57% higher rate of fabricated references
  • Approximately 147,000 AI-generated non-existent citations were estimated in 2025
  • Errors more common in AI-intensive fields and AI-assisted manuscripts
  • Early-career researchers and smaller teams are more frequently affected
  • Hallucinated citations often appeared credible and difficult to detect
  • Most affected papers remained uncorrected or unretracted

Together, the findings highlight a growing challenge: fabricated and AI-generated citations are entering scientific literature at scale, while current peer review systems struggle to detect them. This threatens not only individual studies but also the reliability of the broader evidence base.

Given that research informs clinical guidelines and policy decisions, inaccurate or fictitious citations can compromise the entire evidence chain. Alarmingly, most affected papers had not been corrected or retracted, underscoring gaps in existing quality control mechanisms.

Based on the findings, the experts emphasize the need for systemic solutions, including the integration of automated reference verification tools into journal workflows, improved transparency in indexing systems, and stronger research integrity frameworks. As artificial intelligence continues to shape academic writing, ensuring the accuracy and authenticity of citations will be essential to maintaining trust in scientific research.
References:
1. Topaz, M., Roguin, N., Gupta, P., Zhang, Z., & Peltonen, L. (2026). Fabricated citations: An audit across 2·5 million biomedical papers. The Lancet, 407(10541), 1779-1781. https://doi.org/10.1016/S0140-6736(26)00603-3
2. Zhao, Z., Wang, Y., Stuart, T., De Vaan, M., Ginsparg, P., & Yin, Y. (2026). LLM hallucinations in the wild: Large-scale evidence from non-existent citations. ArXiv. https://arxiv.org/abs/2605.07723
Tags:    
Article Source : The Lancet, arXiv

Disclaimer: This website is primarily for healthcare professionals. The content here does not replace medical advice and should not be used as medical, diagnostic, endorsement, treatment, or prescription advice. Medical science evolves rapidly, and we strive to keep our information current. If you find any discrepancies, please contact us at corrections@medicaldialogues.in. Read our Correction Policy here. Nothing here should be used as a substitute for medical advice, diagnosis, or treatment. We do not endorse any healthcare advice that contradicts a physician's guidance. Use of this site is subject to our Terms of Use, Privacy Policy, and Advertisement Policy. For more details, read our Full Disclaimer here.

NOTE: Join us in combating medical misinformation. If you encounter a questionable health, medical, or medical education claim, email us at factcheck@medicaldialogues.in for evaluation.

Our comments section is governed by our Comments Policy . By posting comments at Medical Dialogues you automatically agree with our Comments Policy , Terms And Conditions and Privacy Policy .

Similar News