The Hidden Risks Of Generative Artificial Intelligence In Accounting And Auditing: A Bibliometric Perspective

Generative artificial intelligence has moved quickly from experimental systems to tools used in routine accounting and audit work. Large language models now assist in drafting audit documentation, summarizing contracts, interpreting unexpected variances and supporting elements of risk assessment. These uses promise efficiency gains, yet they also raise questions about evidence quality, professional judgement and the reliability of financial reporting. This paper examines how recent academic and professional literature addresses these issues. Based on a bibliometric analysis of 286 Web of Science publications (October 2025) that link generative AI with accounting, auditing or financial reporting, the study maps the structure of this fast-growing field. Using VOSviewer keyword co-occurrence analysis, frequency counts and close reading of influential works, the paper identifies four recurring groups of risks: persuasive but weakly evidenced AI-generated outputs, pressures on professional judgement including automation bias, data security and confidentiality vulnerabilities, and ethical or reputational biases embedded in training data. The results show a sharp rise in publications after 2023 and a pronounced interdisciplinary profile, with contributions from accounting, computer science, engineering and education. By connecting these risks to core concepts such as verifiability, audit quality and the public interest, the paper argues that generative AI alters the conditions under which professional judgement is formed and evaluated. The conclusion outlines implications for accounting education, organizational AI policies and future research on human–AI interaction and validation of AI-generated evidence. For practitioners, the framework can also serve as a simple checklist when deciding how and where to integrate generative AI into everyday accounting and audit work.