Artificial Intelligence in Real-Time Payment Systems: A Bibliometric and Thematic Analysis of Emerging Research Trends

Abstract

The rapid diffusion of real-time payment systems (RTPs) has compressed payment authorization, clearing, and fraud-control decisions into near-instantaneous processes, increasing demand for artificial intelligence (AI)-enabled risk detection capable of operating under severe latency, scalability, and security constraints. Although research on digital payments, financial fraud, blockchain, and AI has expanded considerably, the intellectual structure and thematic development of scholarship specifically connecting AI with real-time and instant payment environments remain insufficiently mapped. This study conducts a bibliometric and thematic analysis of 545 documents published between 2004 and 2026, using outputs generated through the Bibliometrix/Biblioshiny environment in R. Performance analysis, source and author productivity, country and institutional contributions, global citation analysis, keyword frequency, trend-topic analysis, co-word mapping, thematic mapping, bibliographic coupling, and international collaboration networks are examined. The corpus records an annual growth rate of 26.29%, while 76.9% of all publications appeared during 2024–2026, indicating a sharply accelerating research domain. Conference papers account for 54.3% of the corpus, reflecting a technology-intensive and rapidly evolving knowledge base. Thematic mapping identifies two principal motor themes: learning systems–fraud detection–machine learning and artificial intelligence–blockchain–security. Graph neural networks and reinforcement learning form a lower-centrality but highly recent research front, while behavioral research, FinTech analytics, risk management, and decision-making constitute a second emerging trajectory. Trend analysis reveals a temporal transition from early interest in instant payments toward blockchain, distributed infrastructure, scalability, and security, followed by a pronounced 2025–2026 shift toward fraud detection, learning systems, and graph neural networks. Bibliographic coupling further identifies three active fronts: operational fraud/anomaly detection, machine/deep-learning-based detection, and blockchain/security architectures. The study concludes that AI research in RTPs is moving from transaction-level classification toward relational, adaptive, privacy-aware, explainable, and operationally deployable intelligence. A future research agenda is proposed around temporal graph learning, concept drift, adversarial robustness, federated learning, explainability, latency-aware benchmarking, cross-institutional datasets, and responsible AI governance.