The Auditor’s Dilemma: AI Adoption, Trust and Regulatory Legitimacy in Auditing

Abstract

Artificial intelligence (AI) is widely expected to reshape auditing, yet its diffusion within the profession remains uneven and conceptually fragmented. Existing studies have examined adoption, analytics, explainability, governance and ethical concerns, but these elements are rarely integrated into a common account of how AI becomes acceptable in audit practice. This paper addresses that gap by examining the literature through the perspective of the auditor’s dilemma: the tension between AI adoption, trust and regulatory legitimacy in a highly regulated profession. It argues that AI in auditing should be understood not merely as a technology adoption issue, but as a broader professional and institutional challenge in which technical capability must be reconciled with explainability, accountability and oversight. Yet these benefits do not remove the human element: auditor experience, professional scepticism and oversight remain essential to interpreting, challenging and validating AI-supported audit evidence (Samiolo et al., 2024). The paper adopts a systematic literature review design guided by a PRISMA 2020-informed approach. Based on a Scopus search, 94 records were identified, screened, and narrowed to a final analytical corpus of 23 studies. The review synthesizes this evidence thematically and develops an integrative conceptual framework. The findings show that AI diffusion in auditing depends on three interrelated conditions. First, adoption requires organizational readiness, including technological maturity, data quality, skills, process compatibility and client readiness. Second, sustained use requires trust-based professional acceptance, supported by explainability, transparency, robustness and compatibility with auditor judgment and professional scepticism. Third, steady dissemination depends on regulatory legitimacy, which is shaped by documentation, traceability, reviewability, governance and compatibility with standards and oversight. The review also shows that AI-enabled procedures may face resistance not only because of technical limitations, but because auditors and regulators may not regard them as legitimate within established audit-quality expectations. This paper contributes in three ways. First, it consolidates a fragmented but rapidly growing body of literature on AI in auditing. Second, it advances an original framework linking organizational readiness, trust-based professional acceptance and regulatory legitimacy. Third, it shifts the debate beyond conventional technology-acceptance reasoning by showing that, in auditing, AI must be not only useful, but also explainable, auditable and institutionally defensible. Overall, the paper concludes that the future of AI in auditing will depend less on technological availability than on whether AI can become trusted, reviewable and regulatorily legitimate within the assurance environment.