We construct magnitude–frequency distributions according to the Gutenberg–Richter (GR) relation for four major active fault earthquake sources used in previous seismic hazard assessments of New Zealand, and evaluate if GR‐based fault models can be considered for future probabilistic seismic hazard analysis (PSHA). Ultimately, we aim to understand whether GR‐based fault models can be used with the characteristic earthquake model, or if they should be removed from consideration. Our GR magnitude–frequency distributions are constructed for the fault sources using constraints from historical earthquakes, paleoseismic data, and assumptions of earthquake clustering. We find that when the resulting distributions are applied to a model that assumes earthquakes are random and independent in time (as in typical PSHA), the model predicts (1) rates of the largest earthquakes that are less than the rates of equivalent earthquakes derived from paleoseismic data, but cannot be ruled out due to the large uncertainties in data and analysis, and (2) rates of earthquakes that are considerably greater than the observed rates. We also undertake epidemic‐type aftershock sequence (ETAS) simulations to determine the potential range of magnitude and frequency that could be expected due to earthquake clustering. We find that the simulations show compatibility between the GR relation and the available constraints for each fault source, which suggests that the GR relation cannot be ruled out as a source of plausible description of the magnitude–frequency distribution for the fault sources. The mixed results of our analysis suggest that including both GR and characteristic earthquake models in PSHA is an appropriate approach at this stage. Although the simultaneous implementation of both models is commonplace in PSHA, we can now justify doing so on the basis of this analysis. Considerations as to the return period of interest in PSHA and the scale dependency of magnitude–frequency distributions may further guide the use of the two models.