Detecting Hallucinated and Suspicious Citations: What Current Tools Can and Cannot Do
TL;DR
It is argued that hallucinated and suspicious references have become a real and growing problem for scientific communication, and that more transparent and multi-source detection systems are still needed.
Abstract
Large language models are increasingly used in academic writing, including for reference generation, raising concerns about hallucinated and unreliable citations. Recent research suggests that this problem is already widespread and is becoming increasingly prevalent in the published literature and at scientific conferences. In this position paper, we review recent studies on hallucinated references and evaluate several currently available tools for detecting problematic references using documents containing hallucinated citations. The tools assessed include CheckIfExist, HalluCiteChecker, Hallucinator, Hallucinated Reference Finder (HalRef), and RefChecker. While these systems can provide useful early warnings in many cases, their performance is limited by reference extraction errors, incomplete metadata, limited database coverage, and inconsistent verification results. We argue that hallucinated and suspicious references have become a real and growing problem for scientific communication, and that more transparent and multi-source detection systems are still needed.