The Growing Problem of Fake Citations in Scientific Research

In the world of science, citations are essential. They act like a family tree, connecting new research to the studies that came before it. By citing previous work, scientists show where their ideas come from and provide evidence for their claims. However, a new study published in The Lancet suggests that this system is facing a serious threat. Researchers are finding an increasing number of "fabricated" citations—references to papers that do not actually exist.

According to STAT News and MedPage Today, experts believe that generative artificial intelligence (AI) tools are likely to blame. These tools, known as large language models (LLMs), are designed to write and edit text. While they are helpful, they are also known for "hallucinating," which means they confidently create information that is not true. In the context of research, these AI tools can invent references that look perfectly real. They often use correct formatting, include the names of real researchers, and list believable publication dates, making them very difficult for human editors to spot.

Maxim Topaz, a researcher at Columbia University who led the study, discovered this problem after a personal experience. While using an AI tool to help edit his own work, he was embarrassed when a journal editor pointed out that one of his citations led to nowhere. This experience inspired him to investigate how widespread the issue had become. His team analyzed millions of biomedical papers and found that the number of fake references has grown significantly. In 2023, about one in 2,828 papers contained a fake citation. By early 2026, that number had jumped to one in 277 papers.

This trend is worrying for many in the scientific community. Misha Teplitskiy, a sociologist of science at the University of Michigan, noted that while people are excited about AI making science faster, there is a lack of focus on the quality of the work being produced. He described this trend as a sign of "slop," or low-quality work. When researchers rely on AI to generate their references without checking them, they risk polluting the public record of science with false information.

Beyond just being a technical error, some experts suggest this problem points to deeper issues in how science is conducted today. Mohammad Hosseini, a professor at Northwestern University who studies research ethics, explained that the rise in fake citations shows that some authors are more interested in publishing quickly than in doing careful work. He argued that the current academic system puts too much pressure on researchers to publish many papers, turning the act of citing other work into a "box-checking exercise" rather than a meaningful part of scientific discussion.

MedPage Today reported that the consequences could be severe. If fake citations become common, they could impact systematic reviews or clinical guidelines, which are documents that doctors use to make important decisions about patient care. In one extreme case, researchers found a paper where 60% of the references were fake. These references were even tailored to look relevant to the specific surgical topic of the paper, showing how sophisticated these AI-generated errors have become.

Topaz warned that his findings are likely just the "tip of the iceberg." As AI tools become more common in the writing process, the risk of these errors will likely continue to grow unless researchers change how they work. The study serves as a warning to the academic community: while technology can help with speed, it cannot replace the human responsibility of verifying facts. For now, the burden remains on authors, editors, and peer reviewers to be more careful, ensuring that the "family tree" of scientific knowledge remains rooted in reality rather than in the imagination of an AI.