AI-Hallucinated Citations Are Quietly Infiltrating the Scientific Record
Researchers warn that AI-generated fake citations are appearing in papers that help shape clinical guidelines. The concern goes beyond embarrassing errors: once incorrect references enter the literature, they can distort evidence reviews and potentially influence care decisions.
The danger posed by hallucinated citations is uniquely serious in medicine because scientific publishing is cumulative. A fabricated or misattributed reference may seem like a small error in a single paper, but it can ripple through reviews, meta-analyses, and guideline development if not caught early.
This is one of the clearest examples of how generative AI can degrade the epistemic quality of healthcare, even when no obvious clinical harm is immediate. In practice, citation hallucinations can waste reviewer time, undermine trust in authors, and create false confidence in literature that was already hard to audit.
The warning also highlights a governance gap. Many institutions have focused on whether AI can draft text or summarize records, but less attention has gone to provenance and citation integrity. For medicine, those details are not administrative trivia; they are the scaffolding of evidence-based care.
The broader lesson is that AI adoption in healthcare cannot be measured only by productivity gains. If the tools accelerate publication without improving verification, they may actually increase the burden on clinicians and researchers who depend on trustworthy evidence.