Study Finds 146,932 Hallucinated Citations in 2025 Research Papers

A team of researchers identified 146,932 hallucinated citations in papers and preprints published in 2025 across four major repositories. The audit reviewed 111 million references from arXiv, bioRxiv, Social Science Research Network, and PubMed Central.

The study found that hallucinated citations appeared at different rates depending on the research field. SSRN, which focuses on social sciences research, showed the highest rate at nearly 2 percent, nearly five times higher than the other repositories examined.

5 million manuscripts and checked them against Semantic Scholar, OpenAlex, and Google Scholar. References that could not be matched and that an LLM judged to be intended as academic sources were flagged as unmatched. The analysis only counted faulty references appearing after 2022, the year ChatGPT became publicly available, to distinguish hallucinations from earlier bibliographic errors.

Yin, assistant professor of information science at Cornell University, said the team was surprised by the scale of the issue. "We were really amazed by the overall magnitude and dynamics of the whole body of hallucinated citations," he said. The study was posted on the preprint server arXiv and has not been peer-reviewed.