Seismogenic depth is a fundamental parameter in seismic hazard assessment and is commonly inferred from kinematic approaches that rely on empirically defined thresholds. However, these observational estimates require validation and calibration against physics-based earthquake cycle models. Here we focus on the San Andreas Fault system in California, where high-quality geodetic, seismicity, and geothermal datasets are available. We construct a geodetically derived fault-coupling model for the entire fault system and systematically compare seismogenic depths inferred from fault coupling with those constrained by earthquake depth distributions. Our results show that a geodetic seismogenic depth defined by a coupling ratio of 0.45 provides the closest agreement with the depth enclosing 90% of the observed seismicity. This correspondence is quantitatively consistent with predictions from thermally constrained rate-and-state friction models, although the numerically inferred seismogenic depths are systematically shallower. Along-strike variations in seismogenic depth obtained from all approaches exhibit similar spatial patterns and correlate strongly with geothermal gradients, indicating that temperature is the primary controlling factor. These results establish a quantitative link between seismogenic depths derived from observational constraints and physics-based numerical models, thereby providing a stronger physical basis for incorporating geodetically inferred coupling models into seismic hazard assessments.