Society & Ethics

AI-hallucinated citations are creeping into papers that shape clinical guidelines, researchers warn

· May 26, 2026
AI-hallucinated citations are creeping into papers that shape clinical guidelines, researchers warn

What happened

An audit of 2.5 million biomedical papers reveals a sharp increase in fabricated citations since 2023. Researchers from Columbia University and partner institutions found that false references are appearing more than twelve times as often as before. These fake citations closely match paper topics and mimic correct formatting, making them extremely hard to detect. The surge is linked to the rising use of AI language models, which tend to produce plausible but nonexistent sources. Most papers with these AI-generated fake citations—98 percent—have not prompted any response from their publishers.

Why it matters

Fabricated citations undermine the trustworthiness of scientific literature, particularly in biomedicine where papers often influence clinical guidelines and patient care protocols. When false references slip into research, they distort the evidence base that doctors, hospitals, and regulators rely on to make decisions. This problem pressures journal editors and publishers to overhaul review and verification processes. For operators in medical research or healthcare, this injects new risks around the reliability of clinical recommendations. It adds a hidden layer of uncertainty that could delay or complicate treatment innovation and policy updates.

What to watch next

Publishers must develop better AI-detection tools to catch fabricated citations before papers are finalized. Watch for advancements in citation verification software and stricter editorial standards targeting AI-generated content. Research institutions and funders may start demanding transparency about how AI tools are used in writing and citing research. Regulators and medical associations could eventually reassess how clinical guidelines are derived to avoid basing them on compromised literature. The problem also raises flags for AI developers on the need to improve model accuracy and reduce hallucination around references.

AI Quick Briefs Editorial Desk

Stay ahead of AI Get the most important AI news delivered to your inbox — free.