At least 50 hallucinated citations found in ICLR 2026 submissions (gptzero.me)

A recent analysis using GPTZero's Citation Check tool uncovered at least 50 instances of hallucinated citations in submissions for the prestigious International Conference on Learning Representations (ICLR) 2026. This issue arises as the influx of papers fueled by generative AI and publication pressure overwhelms the peer review process. Despite being evaluated by multiple expert reviewers, many submissions containing these inaccuracies might have been inadvertently accepted, which poses serious ethical concerns within the academic community. The findings highlight the increasing challenge of maintaining the integrity of scholarly publishing as AI tools like LLMs contribute to the proliferation of "AI slop" — submissions that may lack rigorous factual accuracy. GPTZero’s Citation Check tool is designed to detect such inaccuracies by flagging citations that cannot be verified online, allowing for a human review to determine their legitimacy. This development underscores the urgency for enhanced tools to assist peer reviewers, as the pressure to publish continues to rise, threatening the validity of research outputs in AI and beyond.