Deep excavations underpin transit-oriented districts, airport-linked complexes, high-rise mixed-use precincts, and underground utility corridors. In these dense settings, wall movements, ground loss, drawdown, and construction sequencing interact with vulnerable urban assets in ways that are difficult to observe and harder to predict. Digital twin technology is increasingly proposed as a response because it links field measurements with continuously updated virtual models for interpretation, forecasting, and intervention. This review synthesizes literature published between 2020 and 2025 on digital twins, geotechnical monitoring, deep excavation risk analytics, and urban digital infrastructure. Using a PRISMA-informed review design, the paper maps how sensing, BIM-GIS integration, physics-based simulation, machine learning, and uncertainty management are combined to support wall deflection control, settlement warning, groundwater management, and protection of adjacent buildings. The review argues that the value of a digital twin lies not in visualization alone but in the creation of a closed monitoring loop in which data, models, thresholds, and response rules are aligned to excavation stages and urban exposure. Two contributions are offered: a lifecycle architecture for twin-enabled excavation monitoring and a maturity-based implementation roadmap for mega urban developments. The synthesis shows that current advances are strongest in risk prediction, virtual sensing, and automated model updating, whereas major limitations remain in data interoperability, uncertainty communication, contractual governance, and scaling from pilot projects to district-level portfolios.