A study led by Columbia University’s nursing professor Maxim Topaz warns that AI-generated errors are increasingly infiltrating biomedical scholarship through fabricated references. Topaz and colleagues audited nearly 2.5 million biomedical papers and 97 million citations indexed in PubMed Central, finding more than 4,000 fabricated references across nearly 3,000 papers. The analysis points to a sharp rise: the team reports fabricated-reference rates climbing more than 12-fold over three years, with early 2026 findings indicating roughly one in 277 papers contains at least one non-existent reference. The findings raise immediate quality-control stakes for editors, peer reviewers, and researchers who increasingly use AI tools for drafting and formatting.