In shallow, tidally dominated regions, overtides and the mean state of the ocean are coupled through their simultaneous generation by nonlinear processes. We present a new method that uses observed overtides (e.g., M4) and mean currents to independently assess the accuracy of mean dynamic topography (MDT) predicted by ocean models. This is useful in regions where no sufficiently long, geodetically referenced sea level records are available for validation of the predicted MDT. We apply the new method to a regional model of the Gulf of Maine/Scotian Shelf region (GoMSS) and a barotropic, higher resolution model focused on the upper Bay of Fundy (UBoF). We first show that the tides and mean circulation predicted by UBoF are in good agreement with observations and a significant improvement over GoMSS. Next, we use UBoF to demonstrate that observed overtides are useful in specifying the bathymetry and parameters of an ocean model. An accurate bathymetry is critical for capturing the dominant nonlinear processes that generate overtides and control the form of MDT in shallow, tidally dominated regions. Finally, we use the observed overtides to argue that the MDT predicted by UBoF is more realistic than the prediction by GoMSS. In the vicinity of headlands, both horizontal advection and bottom friction in UBoF generate harmonics of the tidal flow and local setdowns of coastal MDT of O(10 cm). The prediction of such features, validated by observed overtides, can provide guidance in future deployments of tide gauges in support of geoid and ocean model validation.