Artificial Intelligence in Tax Auditing: A Bibliometric and Thematic Analysis

Tax noncompliance continues to be a significant challenge for tax administration. To address this issue, many have started using artificial intelligence and machine learning to enhance how audit cases are selected and fraud is detected. Although individual applications show considerable promise, there has been little effort to systematically map how AI is being used in tax auditing. This study conducts a bibliometric and thematic analysis of 578 Scopus-indexed publications from 1990 to 2025. The study traces the development of the field using several bibliometric techniques. These include performance indicators, co-authorship networks, keyword co-occurrence analysis, co-citation analysis, and bibliographic coupling. The analysis shows three main patterns: output has grown sharply since 2015; the knowledge base is hybrid, combining public finance and computer science; and research fronts are fragmented but thematically focused. These clusters were dominated by work on value-added tax (VAT) and e-invoicing anomaly detection, risk-based audit selection, and compliance or fraud detection. These themes reflect the shift from rule-based auditing to ML and graph-based approaches. While United States produces the most highly cited work, China leads in volume, highlighting an uneven global distribution of influence and practice. The results highlight significant guidance for policy and practice, underlying the need for building audit models that ensure equity and welfare as well as emphasize the need for cross-jurisdictional collaboration and standardized evaluation protocols. Using Scopus as the only data source is a limitation, and future studies would benefit from using multiple sources. Aside from this limitation, this study provides the first field-specific bibliometric map focused specifically on AI in tax auditing, giving researchers, policymakers, and administrators seeking sustainable and data-driven tax enforcement systems a clear picture of current developments, existing gaps, and directions for future studies.