Devansh Gupta, Priyanka Chugh, Poonam Mahajan
As financial fraud becomes more sophisticated and financial services are increasingly digitized, artificial intelligence (AI) and machine learning are emerging as pivotal technologies for risk management and compliance. While research into AI-driven fraud detection is advancing rapidly, the intellectual structure and theoretical underpinnings remain fragmented. This paper provides a systematic review of 118 peer-reviewed articles published between 2015 and 2025, combining bibliometric science mapping with the SPAR-4-SLR protocol to ensure rigour, transparency and replicability. Through co-word network analysis, thematic mapping and conceptual clustering, the study traces the field’s evolution from rule-based systems to adaptive anomaly detection, explainable AI and compliance models, with a focus on digital payment ecosystems and blockchain-enabled applications. The analysis highlights key theoretical anchors, including Fraud Triangle Theory, Agency Theory, Game Theory, Trust and Signalling Theories and regulatory compliance perspectives. It also identifies underexplored areas such as federated learning, algorithmic auditing and cross-jurisdictional intelligence. By mapping theoretical foundations and thematic development, this study offers an evidence-based account of how AI in fraud detection has evolved. It concludes by proposing a future research agenda emphasizing transparency, ethical assurance and global governance alignment, advancing financial risk management through conceptual clarity, methodological guidance and actionable pathways for responsible AI adoption.