Bridges constitute critical links within transportation networks. Therefore, their timely maintenance is essential for safety and sustainability. Data-driven, AI-based decision-support tools are promising for optimising maintenance planning by learning from data and simulating long-term outcomes, often outperforming traditional approaches. Yet their practical adoption remains limited as they lack interpretable decision-making logic, reducing trust, complicating validation, and slowing integration into established workflows. This study introduces an interpretable reinforcement learning (RL) framework to support transparent maintenance decision-making for bridges within a broader road network. The framework integrates data-based degradation and maintenance-effect models to predict asset states over time, enabling the agent to learn maintenance strategies that balance cost and long-term performance. The architecture is structured to reflect human expert reasoning, ensuring model decisions remain interpretable. For example, the framework explores hierarchical learning structures to decompose complex optimisation tasks into recognisable sub-decisions, such as when to perform maintenance or which asset type to prioritise. The network’s graph structure is also exploited to model asset conditions and interactions, allowing the policy to account for how each action affects both individual assets and the broader system. This architecture enables the policies to retain the advantages of RL while remaining accessible and traceable for human experts. A case study based on data-derived deterioration and maintenance scenarios demonstrates the approach’s feasibility in a simulated environment. The developed algorithm combines predictive modelling and interpretable RL architectures within a coherent decision-support framework, addressing a key implementation challenge in AI-supported infrastructure maintenance planning. In doing so, this work contributes to the development of explainable, reliable, and sustainable decision-support systems for bridge and road network maintenance.