A Systematic Review of Explainable AI in B5G/6G Wireless Networks

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The integration of Artificial Intelligence (AI) into Beyond 5G (B5G) and 6G wireless networks has led to an increasing use of Machine Learning (ML) models across different network functions, including Radio Access Networks (RAN), network slicing, beam management and cybersecurity. However, the architectural complexity of these models has led to concerns regarding transparency, reliability, and trust in highly dynamic and time-critical network environments. Consequently, eXplainable Artificial Intelligence (XAI) has emerged as a potential solution to improve the understanding of AI-driven decisions in future wireless systems. This paper presents a Systematic Literature Review (SLR) of XAI approaches applied in B5G/6G wireless networks leveraging Kitchenham SLR methodology along with PRISMA reporting guidelines. A total of 88 papers comprised of peer-reviewed studies, grey literature, and arXiv preprints were analysed using four quality assessment criteria covering relevance to 6G, technical implementation depth, evaluation rigor, and reproducibility. Studies meeting the predefined quality threshold were retained for synthesis. The findings show that current research is widely dominated by post-hoc explainability approaches, particularly SHAP (34%) and Hybrid SHAP-LIME (22%), while intrinsic explainability techniques remain comparatively less explored (4% or less). Although many studies demonstrate the feasibility of integrating XAI into 6G-related applications, several important limitations remain insufficiently addressed, including computational overhead, latency, scalability, robustness of explanations and the impact of feature dependencies on explanation reliability. Additionally, most experiments are conducted within controlled or small-scale experimental settings which do not adequately represent the complexity, heterogeneity and ultra-low latency requirements expected in 6G. Overall, the review highlights a growing gap between current research and practical deployment requirements, suggesting that substantial research work is required before XAI can be reliably integrated into real-time AI-native future wireless systems.