Study Finds AI Hallucinations Boost Fake Citations
SkimNews Take
The rapid proliferation of AI-generated content is transforming the very definition of "original research" by introducing unverified information into foundational academic records.
Get the Health newsletter
Daily health & science — research, biotech, public health, the studies worth knowing. Free.
- Columbia University researchers analyzed over 2 million papers and 97 million citations with AI tools, uncovering about 4,000 fabricated references in 2,800 papers.
- Fabricated citations rose sixfold, from 1 in 2,828 papers in 2023 to 1 in 458 in 2025, and reached 1 in 277 papers during the first seven weeks of 2026.
- Maxim Topaz reported that an AI chatbot introduced a non‑existent citation into his own manuscript, prompting the study’s focus on AI‑induced hallucinations.
- White House MAHA report on chronic disease contained several incorrect citations that observers suspect were generated by AI, illustrating the problem beyond academia.
- Misha Teplitskiy warned that AI‑generated content is creating “slop” in scientific literature, threatening the reliability of systematic reviews and clinical guidelines.
Why it matters: Researchers and journals lose credibility while funding agencies and AI developers face stricter verification standards, as the rate rose to 1 in 277 papers, a sixfold increase, jeopardizing systematic reviews and clinical guidelines.
