Study finds nearly 147,000 AI-generated fake citations in scientific papers across major research databases

A groundbreaking study has exposed nearly 147,000 AI-generated fake citations embedded within scientific papers across leading academic databases. This unprecedented surge in erroneous references is attributed to researchers misusing large language model (LLM) based AI chatbots, failing to verify the legitimacy of generated references before inclusion. Concurrently, arXiv, a pivotal open-access preprint server, has formally announced a ban on submissions containing such 'hallucinated citations,' signaling a critical pivot in safeguarding academic integrity. This alarming discovery underscores a significant challenge to the scientific publishing ecosystem and broader information integrity. In an era of intense academic competition and the 'publish or perish' paradigm, the rapid proliferation of Generative AI tools without commensurate ethical guidelines or rigorous verification protocols creates dangerous shortcuts. The erosion of trust in scholarly output could have cascading effects, from misinformed policy decisions and flawed subsequent research to a general degradation of public confidence in science. This incident highlights the urgent need for enhanced digital literacy, more robust peer review processes, and clear institutional policies to navigate the dual-edged sword of AI in research.