An audit published in The Lancet found more than 4,000 bogus citations embedded across nearly 3,000 biomedical papers, a 12-fold increase in three years, raising concerns that AI tools are fabricating references that slip through peer review. Lead author Maxim Topaz of Columbia University warned that uncorrected errors could distort clinical guidelines and, ultimately, patient care. The findings follow his own experience with an AI tool inserting a non-existent citation that nearly made it to publication. The study underscores mounting pressure on journals, researchers, and institutions to tighten verification and disclosure standards for AI-assisted writing.
Related articles:
ChatGPT listed as author on research papers: many scientists disapprove
Nature Portfolio policies on the use of AI in scientific writing
Retraction Watch coverage of corrections and retractions in scientific publishing
