A landmark study examining AI-generated citations in biomedical literature has revealed that artificial intelligence systems may have been systematically fabricating not just individual papers, but entire academic disciplines, researchers reported in a paper that is already drawing significant attention from the scientific community.

The study, which examined 847 AI-generated citations over a period of eighteen months, found that computational models appear to have independently developed coherent theoretical frameworks for fields including "Therapeutic Nostalgia," "Quantum Bedside Manner," and "Evidence-Based Longing." Dr. Marieke Van Der Berg, Professor of Computational Bibliography at the Karolinska Institute and lead author of the study, expressed bewilderment at the findings. "What's particularly striking is that these fabricated citations follow proper academic formatting conventions while describing research methodologies that may not technically exist," Van Der Berg told reporters.

The implications of these findings could reshape how the scientific community approaches interdisciplinary collaboration, particularly given that several research institutions have reportedly begun receiving grant applications for studies in these AI-invented fields. Dr. James Thornfield, Associate Professor of Digital Epistemology at Oxford University and not involved in the current study, suggested the problem may be more widespread than initially suspected. "We're seeing evidence that AI systems may have been cross-referencing their own fabricated citations, creating self-reinforcing academic ecosystems," Thornfield explained.

What makes this finding particularly fascinating is that the AI-generated papers appear to maintain internal logical consistency, with one fabricated study building methodologically upon another fabricated study in ways that mirror legitimate scientific discourse. The research team noted that at least three major universities have established reading groups dedicated to discussing these non-existent papers.

"The real question," Van Der Berg told reporters, "is whether we're prepared to accept that AI may understand academic rigor better than academics do."