Apostolos Dasilas

Department of Applied Informatics

University of Macedonia

E-mail: dasilas@uom.edu.gr

Margarita Trachana*

Department of Applied Informatics

University of Macedonia

E-mail: mtrachana@uom.edu.gr

* Corresponding author

Extended Abstract

Peer-to-peer (P2P) lending has emerged as a significant financial innovation over the past decade, enabling direct interaction between borrowers and investors while bypassing traditional financial intermediaries. This approach offers lower operational costs, faster loan processing, and greater financial accessibility. However, the sector faces substantial risks, including borrower default, fraudulent activities, and market instability. These risks are shaped by platform architecture, regulatory environments, organizational structures, transaction types, and stakeholder interactions. Artificial Intelligence (AI) presents promising opportunities to mitigate such risks through advanced credit scoring, fraud detection, and predictive analytics. Nevertheless, existing AI-based solutions often focus excessively on algorithmic accuracy while neglecting business-context integration, explainability, and regulatory compliance.

This study conducts a Systematic Literature Review (SLR) combined with a bibliometric analysis to examine AI applications for risk mitigation in P2P lending. The review evaluates how machine learning, deep learning, and other AI models have been deployed to enhance platform resilience, analyses the factors influencing platform success or failure, and identifies research gaps limiting real-world applicability. The PRISMA framework was adopted to ensure methodological rigor. Publications from 2015 to 2025 were retrieved from Scopus and Web of Science using targeted keyword searches. Eligible studies underwent qualitative synthesis to categorize methodological approaches and business considerations, while bibliometric mapping was conducted using VOSviewer and R-programming to visualise co-authorship networks, keyword co-occurrence, and thematic trends.

The results reveal a substantial increase in AI-focused P2P lending research from 2018 onwards, with China, the United States, and the United Kingdom emerging as leading contributors. Three major thematic clusters were identified: (1) predictive credit risk modeling using supervised learning algorithms, (2) AI-enabled fraud detection through anomaly detection and pattern recognition, and (3) hybrid frameworks combining AI with traditional financial indicators. Although these approaches achieve strong predictive performance, they frequently lack integration with operational and contextual variables, fail to address model interpretability, and rarely evaluate cross-platform adaptability. Ethical considerations, regulatory alignment, and the impact of AI adoption on user trust are also underexplored.

The review concludes that AI holds substantial potential to improve the robustness and efficiency of risk management in P2P lending. However, future research should focus on the development of explainable AI models to enhance transparency and stakeholder trust, adopt context-aware frameworks incorporating business and market variables, and ensure compliance with regulatory frameworks to protect both investors and borrowers. Achieving these goals will require interdisciplinary collaboration between data scientists, finance professionals, regulators, and platform operators.

By consolidating existing knowledge and mapping current research trends, this study provides a structured foundation for advancing AI-driven risk mitigation strategies in P2P lending. The findings offer actionable guidance for researchers pursuing practically relevant studies, policymakers developing informed regulations, and platform operators seeking to implement ethical, transparent, and effective AI systems. This work contributes to the sustainable evolution of P2P lending as a credible and resilient alternative to traditional financial intermediation.