Temporally and spatially clustered earthquake sequences along plate boundary zones indicate that patterns of seismicity may be influenced by earthquake-induced stress changes. Many studies invoke Coulomb stress change (CSC) as one possible geo-mechanical mechanism to explain stress interactions between earthquakes, their aftershocks, or large subsequent earthquakes; however, few address the statistical robustness of CSC triggering beyond spatial correlations. To address this, I evaluate the accuracy of CSC predictions in subduction zones where Earth’s largest earthquakes occur and generate voluminous and diverse aftershock sequences. A series of synthetic tests are implemented to investigate the accuracy of inferred stress changes predicted by slip distributions inverted from suites of geodetic observations (InSAR, GPS, seafloor geodetic observations) that are increasingly available for subduction zone earthquakes. Through these tests, I determine that inferred stress changes are accurately predicted at distances greater than a critical distance from modeled slip that is most dependent on earthquake magnitude and the proximity of observations to the earthquake itself. This methodology is then applied to the 2010 Mw 8.8 Maule, Chile earthquake sequence to identify aftershocks that may be used to perform statistically robust tests of CSC triggering; however, only 13 aftershocks from a population of 475 events occurred where confidence in CSC predictions is deemed to be high. The inferred CSC for these events exhibit large uncertainties owing to nodal plane uncertainties assigned to the aftershock mechanisms. Additionally, tests of multiple published slip distributions result in inconsistent stress change predictions resolved for the 13 candidate aftershocks. While these results suggest that CSC imparted by subduction megathrust earthquakes largely cannot be resolved with slip distributions inverted from terrestrial geodetic observations alone, the synthetic tests suggest that dramatic improvements can be made through the inclusion of near-source geodetic observations from seafloor geodetic networks. Furthermore, CSC uncertainties will likely improve with detailed earthquake moment tensor catalogs generated from dense regional seismic networks.