Auditing at the Algorithmic Frontier: A Critical Narrative Review of Machine Learning and Audit Quality

Abstract

The rapid integration of machine learning (ML) and artificial intelligence (AI) into audit practice has generated growing scholarly interest in their implications for audit quality. This narrative review synthesizes 49 peer-reviewed articles sourced from the Web of Science (WoS) database, spanning 2010 to 2026, to critically examine how ML adoption shapes audit quality across diverse institutional and organizational contexts. Drawing on institutional theory and the socio-technical systems framework, this study proposes an integrative conceptual framework that positions five independent variables: institutional pressures, technological capabilities, strategic orientation, ethical frameworks, and AI autonomy level, as antecedents of audit quality, mediated by trust, legitimacy, and decision rights, and moderated by governance mechanisms, organizational culture, and institutional environment. The review reveals that whilst ML meaningfully enhances misstatement detection, risk stratification, and processing efficiency, persistent concerns remain regarding auditor over-reliance, algorithmic opacity, accountability displacement, and professional de-skilling. These findings carry significant implications for standard-setters, audit practitioners, and researchers seeking to govern AI responsibly within the auditing profession.