Cultural heritage masonry structures are highly susceptible to various failure mechanisms, such as distortions induced by differential settlements, cracking from seismic activity, and other age-related degradation. These structures often possess unique architectural and material properties, which makes their assessment complex, demanding highly specialized approaches. This paper introduces a novel methodology for risk assessment using Bayesian Belief Networks (BBNs), specifically designed for cultural heritage masonry structures at the territorial level. By employing BBNs, the approach integrates diverse data sources, such as historical records, sensor data, and expert knowledge, to model and quantify the probability associated with different failure scenarios under various conditions for a specific area. The methodology allows for an assessment that is not only data-driven but also flexible, accommodating uncertainty and incorporating dependencies among different failure mechanisms. Using specific case studies and calibrated digital twins (DT), the objective is to demonstrate the framework’s effectiveness in evaluating risks associated with differential settlements induced by ground deformations, excavations or earthquakes, as well as cracking, and progressive deterioration, providing valuable insights for heritage preservation and preventive intervention planning. This approach highlights the potential of BNs to improve risk-informed decision-making, enabling more robust protection of cultural heritage structures.