Abstract Previously hypothesized fault weakening mechanisms include faults lined by low-friction clay gouges, elevated pore pressures within fault cores and/or the operation of dynamic weakening during seismic slip. Geological studies to support dynamic weakening are still in their infancy and there is little geological evidence for the widespread occurrence of low-friction gouges. The cores of some ancient faults exhumed from <5 km depth contain sheared syntectonic mineral veins. This observation is consistent with elevated pore pressures, but the implications for long-term fault weakening are unclear. Experimental data and microphysical modelling suggest that frictional–viscous flow within phyllosilicate-rich fault rocks (phyllonites, some foliated cataclasites) can cause sufficient weakening of crustal faults to satisfy published heat flow constraints. These predictions are consistent with the common occurrence of phyllonite in the cores of large-displacement faults exhumed from >5 km depth. Comparison with seismological data suggests that some faults with phyllonitic cores are likely to generate large earthquakes. Future studies should establish the geological evidence for seismic slip within phyllonitic fault cores and quantify the partitioning between seismic slip and frictional–viscous flow. Further geological observations are also required to test the hypothesized mechanisms by which earthquakes can nucleate and propagate along phyllosilicate-rich faults.