Anna Hartz
Thousands of scientific papers contain references to studies that never existed.
Artificial intelligence has become a common tool in research, education, and everyday work. It can save time and help people process information quickly. But experts continue to warn that AI systems can sometimes create false information that looks convincing.
Steep incline in fake references
A new study has raised concerns about how this problem is affecting scientific research. Researchers found that thousands of medical papers contain references to studies that do not actually exist, reports Videnskab.
The study, published in The Lancet, examined 2.5 million biomedical articles. Researchers identified more than 4,000 fabricated references spread across 2,810 papers. According to the team, the number of false citations has increased sharply since 2023.
In early 2023, researchers found around four fake references for every 10,000 articles. By the beginning of 2026, that number had climbed to nearly 57 per 10,000 papers.
The researchers believe the rise may be linked to the growing use of AI tools such as ChatGPT, Claude, Gemini, and Copilot. These systems are known to occasionally invent information. Scientists say some cases may involve deliberate fraud, while others could be the result of careless use of AI.
A serious issue
Professor Tove Godskesen, who studies ethics in healthcare and research, said the issue becomes serious when scientists fail to verify their sources before publishing.
She warned that false references could eventually affect decisions made by doctors and patients. If studies rely on non-existent sources, mistakes could spread through the medical community.
The research team found that review papers were especially affected. These articles, which summarize previous research, contained significantly more false references than regular studies.
Another concern is that fake citations can continue to spread. Once they appear in one article, they may be copied by others and gradually gain credibility despite having no basis in reality.
Almost none of the papers identified in the study have been corrected or withdrawn.
Experts are now calling on scientific journals and AI companies to do more to prevent the problem. At the same time, researchers themselves are being reminded that checking sources remains an essential part of scientific work.
