Study Finds Fraudulent Citations Increasingly Common in Medical Research Papers

A May 7, 2026 Lancet study by Columbia University researchers found fabricated citations spreading in medical literature, identifying around 4,000 fake references among 2,800 papers. The rate accelerated from 1 in 2,828 papers in 2023 to 1 in 458 in 2025, reaching 1 in 277 during the first seven weeks of 2026. Researchers attribute the problem to generative AI hallucinations, with more than a third of fabricated citations coming from two large open-access publishers.

Fabricated citations in medical research

• ▪Columbia University researchers identified around 4,000 fabricated citations among 2,800 papers by analyzing over 2 million papers and 97 million citations
• ▪A study published on May 7, 2026 in the Lancet found that fabricated citations that do not reference real papers are spreading in medical research literature
• ▪Columbia University researchers published a dashboard alongside their May 7, 2026 Lancet paper showing the distribution of fabricated citations across publishers
• ▪More than a third of fabricated citations come from two large open-access publishers

AI-generated reference hallucinations

• ▪Columbia University researchers concluded that tools using generative AI are likely to blame for the spread of fabricated citations in medical research
• ▪Maxim Topaz, a nurse and health AI researcher at Columbia University, led the study on fabricated citations published in the Lancet on May 7, 2026
• ▪A White House MAHA report issued in 2025 on priorities in tackling chronic disease contained several incorrect citations suspected to be generated with artificial intelligence
• ▪Maxim Topaz used an AI chatbot to help edit an editorial and an erroneous citation was flagged by a journal editor despite Topaz checking for accuracy

Rising frequency of citation errors

• ▪During the first seven weeks of 2026, the rate of fabricated citations reached 1 in 277 medical research papers
• ▪In 2023, 1 in 2,828 medical research papers contained one or more fabricated references
• ▪In 2025, 1 in 458 medical research papers contained one or more fabricated references, representing a sixfold increase in frequency from 2023

Impact on research quality

• ▪Mohammad Hosseini, a professor at Northwestern University who studies research integrity, stated that the presence of hallucinated citations shows researchers do not want to spend time checking references and reflects desperation to get published
• ▪Mohammad Hosseini stated that citation practices are changing with generative AI use, with researchers using hunches to prompt AI tools for citations rather than engaging in a reflective process of reading and noting papers
• ▪Misha Teplitskiy, a sociologist of science at the University of Michigan, stated that the Lancet study is one of the first papers telling us something about the quality of what is being produced with large language models and is a signal of slop

Publisher validation tools

• ▪The Science family of journals uses an automated tool to check references and has not yet had an issue with a published paper having a fabricated citation
• ▪Renee Hoch, head of publication ethics at PLOS, stated that PLOS has seen numerous unverifiable references in submissions and is exploring options to incorporate reference validation into publishing workflows
• ▪The New England Journal of Medicine and JAMA have tools to validate citations

Journal editorial responses

• ▪Renee Hoch stated that PLOS has seen a lot of false positives in piloting reference validation options, with legitimate references flagged due to erroneous information, formatting issues, publication language, or limitations of research literature databases
• ▪Authors publishing in the New England Journal of Medicine and JAMA agree to be held responsible for their work