Concerns around the reliability of scientific research are growing as artificial intelligence tools become more widely used by researchers. A recent investigation has found that fake citations created by AI systems are slipping into papers accepted at leading global conferences.
According to a report by AI detection startup GPTZero, more than 100 hallucinated citations were identified across 51 research papers accepted at NeurIPS 2025. The company said it scanned 4,841 papers presented at the annual AI and machine learning conference held in December last year in San Diego, California, United States. While 51 papers may not appear statistically significant, NeurIPS policy treats any hallucinated citation as valid grounds for rejection or revocation.
“These NeurIPS papers have already been accepted, presented live and effectively published. Since NeurIPS 2025 had an acceptance rate for main track papers of 24.52%, each of these papers beat out 15,000 other papers despite containing one or more hallucinations,” GPTZero said in a blog post published on Wednesday, January 21. The findings highlight that even experienced AI researchers can struggle to verify outputs from the tools they use.
The issue is not limited to NeurIPS. GPTZero also detected over 50 hallucinated citations in papers under review for ICLR 2026 in December last year. Online pre print platforms have seen a surge in low quality AI generated research. An analysis cited by a global magazine found that researchers using large language model tools posted around 33% more papers than those who did not.
GPTZero said it used its in house agentic AI tool called Hallucination Check to scan cited sources and flag references that could not be found online. Each flagged citation was then verified manually by a human reviewer. The company refers to such references as “vibe citations.” “We define a vibe citation as a citation that likely resulted from the use of generative AI. Our definition excludes obvious spelling mistakes, dead URLs, missing locators and other errors that are plausibly human,” it said. NeurIPS, founded in 1987, has grown rapidly, with attendance reaching 26,000 last year, according to a business news channel. Submissions rose over 220% between 2020 and 2025, putting added pressure on peer review systems. Ironically, a paper titled “The AI Conference Peer Review Crisis,” published months before NeurIPS 2025, had already warned about the risk of AI generated fake citations.
Also read: Viksit Workforce for a Viksit Bharat
Do Follow: The Mainstream formerly known as CIO News LinkedIn Account | The Mainstream formerly known as CIO News Facebook | The Mainstream formerly known as CIO News Youtube | The Mainstream formerly known as CIO News Twitter
About us:
The Mainstream is a premier platform delivering the latest updates and informed perspectives across the technology business and cyber landscape. Built on research-driven, thought leadership and original intellectual property, The Mainstream also curates summits & conferences that convene decision makers to explore how technology reshapes industries and leadership. With a growing presence in India and globally across the Middle East, Africa, ASEAN, the USA, the UK and Australia, The Mainstream carries a vision to bring the latest happenings and insights to 8.2 billion people and to place technology at the centre of conversation for leaders navigating the future.
