Artificial Intelligence (AI) has emerged as a revolutionizing solution in tax administration. However, little research has investigated the trends in using AI to detect tax fraud. This study employs a bibliometric approach. It collected 371 publications indexed in Scopus from 1994 to 2024 and deployed content-based analysis methods to further examine the research content. The content analysis shows four main research themes: machine learning, big data analytics, data mining and fraud prediction, and security and privacy assurance, which are the prominent current research trends. The results also indicate a significant growth in the number of publications, especially since 2019, as AI applications have increasingly been adopted in data analysis. The research collaboration network chart identifies India, the United States, and China as the countries with the most publications and research collaborations with multiple nations worldwide. This study is among the very first papers to combine bibliometric and content analysis with a focus on AI application trends in tax fraud detection.