Seismic structural monitoring requires rapid adaptation of sensing and computation while confirming that uncertain measurements or inconsistent digital-twin (DT) outputs are not released as reliable information. This study proposes a confidence-gated seismic DT-driven cyber-physical system (CPS) framework that integrates event-triggered operation, edge-level data assurance, DT updating, and governed service release. Time-critical functions, including timestamp validation, buffering, data-completeness assessment, synchronization, event confirmation, and hard-constraint screening, are performed close to the data source. Higher-fidelity state estimation, model updating, persistent storage, and cross-event analysis are allocated to digital-twin and cloud resources according to their timing and computational requirements. A provisional system-level confidence index combines signal quality, data completeness, synchronization, physical-digital consistency, and latency adequacy. When integrated with hard-constraint checks, the index classifies each output as trusted, requiring engineering review, or fail-safe. The major contribution is the decoupling of event recognition from update permissibility, which avoids a detected event from being automatically regarded as a reliable structural state. The proposed framework also presents layered architecture, computational workflow, audit criteria, and staged evaluation pathway. Calibration and validation using field-monitoring data or real-time hybrid simulation are required before any automated feedback function can be authorized.