The Role of Statistical Concepts in the Development of Artificial Intelligence for Auditing: A Systematic Literature Review

Abstract

Artificial Intelligence (AI) has increasingly been adopted in auditing, however the role of statistical methods in improving AI performance and reliability remains fragmented. This study aims to systematically identify, review, and synthesize the role of statistical concepts in AI development for auditing, including their implementation, benefits, challenges, and future directions. A Systematic Literature Review (SLR) following the PRISMA framework was conducted on 149 studies published between 2017 and 2026 from Google Scholar, Scopus, and SciSpace. The findings reveal a significant increase in AI auditing research since 2023. Regression analysis, hypothesis testing, and Bayesian inference are the most frequently applied statistical methods, supporting fraud detection, risk assessment, audit sampling, anomaly detection, and model validation. Integrating statistical methods with AI improves prediction accuracy, interpretability, transparency, and audit quality. However, challenges remain regarding data quality, model interpretability, auditor competency, and AI governance. Future research should prioritize hybrid AI-statistical models, Explainable AI, and adaptive Bayesian approaches to enhance trustworthy data-driven auditing