With the advancement of mobile AR technologies, a variety of outdoor Location-Based Augmented Reality (LAR) application systems using smartphones have emerged. They use street-level open 3D geometry data to distinguish outdoor structures, such as buildings, terrain, and walkable navigation paths. Then, for inspecting its spatial accuracy, it is important to estimate user-to-geometry distances accurately in real time without LiDAR sensors. In this paper, we investigate a user-to-building distance estimation method using a mobile visualization tool. It integrates cloud-based street-level geometry data with real-time semantic 3D bounding boxes of buildings and walkable areas. For evaluations, we compared estimated distances against ground-truth measurements by a laser rangefinder across diverse outdoor environments with varying geometric complexity. The results demonstrate that the estimation error remains within 0.9m, with consistently aligned bounding boxes relative to physical structures.