The role of subduction zone geometry in the nucleation and propagation of great-sized earthquake ruptures is an important topic for earthquake hazard, since knowing how big an earthquake can be on a given fault is fundamentally important. Past studies have shown subducting bathymetric features (e.g. ridges, fracture zones, seamount chains) may arrest a propagating rupture. Other studies have correlated the occurrence of great-sized earthquakes with flat megathrusts and homogenous stresses over large distances. It remains unclear, however, how subduction zone geometry and the potential for great-sized earthquakes (M 8+) are quantifiably linked—or indeed whether they can be. Here, we examine the potential role of subduction zone geometry in limiting earthquake rupture by mapping the planarity of seismogenic zones in the Slab2 subduction zone geometry database. We build from the observation that historical great-sized earthquakes have preferentially occurred where the surrounding megathrust is broadly planar, and we use this relationship to search for geometrically similar features elsewhere in subduction zones worldwide. Assuming geometry exerts a primary control on earthquake propagation and termination, we estimate the potential size distribution of large (M 7+) earthquakes and the maximum earthquake magnitude along global subduction faults based on geometrical features alone. Our results suggest that most subduction zones are capable of hosting great-sized earthquakes over much of their area. Many bathymetric features previously identified as barriers are indistinguishable from the surrounding megathrust from the perspective of slab curvature, meaning that they either do not play an important role in arresting earthquake rupture or that their influence on slab geometry at depth is not resolvable at the spatial scale of our subduction zone geometry models.