According to recent findings, artificial intelligence is falsifying references to medical studies that do not exist.
According to an article in The Lancet, a recent audit found that out of millions of biomedical papers, more than 4,000 cited non-existent research. Such fabricated citations could undermine the clinical guidelines that health care professionals rely on to provide care, said Maxim Topaz, an associate professor at Columbia School of Nursing and lead author of the study.
An audit of millions of biomedical papers found more than 4,000 citations of false studies, researchers said in a recent article in The Lancet.
fabricated citations Maxim Topaz, associate professor at Columbia School of Nursing and lead author of the study, told CBS News these guidelines are dangerous because they influence clinical guidelines, which are based on public research that health care professionals follow when providing care.
“When these false references are written into the literature, they end up in these guidelines, and that’s how doctors decide how to provide care for you,” he said. “Your doctor may be making treatment decisions based on research that never existed.”
a growing problem
Equally troubling, he said, is the fact that none of the errors Topaz and his team discovered have been corrected or retracted and could still impact patient care.
“The proportion of false citations appearing in published medical literature is increasing,” Topaz added, noting that the number of such miscitations has increased 12-fold in the past three years. These forged references cover nearly 3,000 academic papers.
Topaz’s own experiences prompted him to investigate the issue. He told CBS News that an artificial intelligence application he used to help improve one of his scientific papers inserted false citations. It then went through several layers of peer review until an eagle-eyed editor spotted the bogus reference.
“I feel ashamed because I’ve been working on artificial intelligence for the past 15 years, so if it could happen to me, it could happen to anyone,” he said.
Topaz explained that such incidents occur when an author asserts a factual statement and asks the AI to cite it. “In some cases, AI will inadvertently shoehorn these things in,” he said. “You hope the facts are accurate, but if they’re backed up by fabricated quotes, you don’t know if the ‘facts’ are accurate.”
In some cases, AI tools also cite real authors when they invent research and attribute it to that person. Other times, Topaz said, the citations were completely fabricated.
“This is just the tip of the iceberg,” he said, noting that research in other areas may encounter the same problem.
At the same time, fake scientific citations generated by AI can “look completely genuine,” Topaz added, stressing the importance of researchers rigorously checking the facts of their work.


