Creeping faults are typically not associated with large earthquakes. However, new K/Ar dating and biomarker maturity data on the San Andreas Fault Observatory at Depth (SAFOD) present evidence that large paleoearthquakes have occurred in the creeping section of the San Andreas Fault, California. K/Ar ages of bulk samples with evidence of coseismic heating range from 3.3 to 15.8 Ma, and argon diffusion experiments suggest that these ages are only partially reset and the actual event ages may be even younger. Thus, questions remain as to how we can refine such dates to reveal the precise age and location of these earthquakes. To refine the ages and more accurately assess seismic hazard, we date size separates of eight samples from different sections of the SAFOD core. Following Stokes’ Law, we split each sample into five size fractions using hydrodynamic settling: <0.2, 0.2-0.5, 0.5-0.8, 0.8-1.4, and 1.4-2 micrometers. The finest size fractions contain the most authigenic illite, which form during fault slip. We determined chemical composition and separated illite polytypes using x-ray diffraction, and also measured K/Ar ages on each sample. Preliminary results from two scaly black fault rock samples, previously shown to have hosted earthquakes, (3,193.69 m and 3,193.96 m along the core) support that the finest size fractions contain the greatest ratio of authigenic illite. With a York regression between age and detrital illite abundance, we place the authigenic illite ages at 1.08 ± 2.40 Ma and 0.88 ± 5.08 Ma for these two samples, and observe that the detrital illite matches the late Cretaceous age for the country rock. This new age estimate for the authigenic illite means that large earthquakes must have propagated into the creeping section within the last million years. Not only is it significantly younger than the bulk sample age, it is recent enough that translation of faulted material from the locked southern San Andreas fault into the creeping section cannot explain the record. Moving forward, we will expand our procedure to include isotope dilution for measuring K concentration and analyze the other samples previously measured for biomarker maturity and bulk K/Ar age. Resulting insights into the fault rock composition and the timing of past earthquakes will be crucial in assessing the region’s seismic hazard.