Whether the stress-loading of faults to failure in earthquakes appears to be random or to an extent explainable, given constraints on fault/shear-zone interaction and the build-up and release of stress over many earthquake cycles, is a key question for seismic hazard assessment. Here we investigate earthquake recurrence for a system of 25 active normal faults arranged predominantly along strike from each other, allowing us to isolate the effects of stress-loading due to regional strain versus across-and along-strike fault interaction. We calculate stress changes over 6 centuries due to interseismic loading and 25 > Mw 5.5 earthquakes. Where only one fault exists across strike, stressloading is dominated by the regional tectonics through slip on underlying shear zones and fault planes have spatially smooth stress with predominantly time-dependent stress increase. Conversely, where faults are stress-loaded by across-strike fault interactions, fault planes have more irregular stress patterns and interaction-influenced stress loading histories. Stress-loading to failure in earthquakes is not the same for all faults and is dependent on the geometry of the fault/shear-zone system. The stress-loading of faults to failure in earthquakes is driven by regional tectonics, but is also influenced by fault interaction during earthquakes, evidenced by calculations of Coulomb stress transfer (CST) and corresponding changes in the rate of seismicity 1-3. Faults also interact over longer time periods evidenced by fault displacement profiles that show steep displacement gradients and enhanced displacements on adjacent fault tips 4,5 , and observations of finite fault displacement profiles that adhere to global scaling relationships between fault length and fault displacement, both for isolated faults and for closely-spaced fault networks (d = γL, where γ = 0.03 for both isolated faults and summed across the strike of fault networks 6,7). However, despite the above evidence for organisation of both the stress-loading to failure process and long-term displacement accumulation, earthquake recurrence is often considered to be a random Poisson process for some seismic hazard purposes 8,9. The link between fault interaction during single earthquakes and over multiple seismic cycles described above argues that the recurrence of earthquakes must sum to produce the long-term displacement profiles. Hence displacement accumulation during earthquakes over the long-term is not random, and must be influenced by interaction, which is in turn governed by fault system geometry. However, study of geometry-controlled interaction is complicated by the fact that fault systems are complex, commonly exhibiting multiple overlapping faults, and it is difficult to isolate the effects of earthquake-inducing stress accumulation caused by tectonic loading as opposed to loading due to rupture of neighbouring faults both along and across strike 10. Without isolation of these effects it is challenging to assess what aspects of the stress-loading to failure, and h...