The Pallett Creek paleoseismic record occupies a keystone position in most attempts to develop rupture histories for the southern San Andreas fault. Previous estimates of earthquake ages at Pallett Creek were determined by decay counting radiocarbon methods. That method requires large samples which can lead to unaccounted sources of uncertainty in radiocarbon ages because of the heterogeneous composition of organic layers. In contrast, accelerator mass spectrometry (AMS) radiocarbon dates may be obtained from small samples that have known carbon sources and also allow for a more complete sampling of the section. We present 65 new AMS radiocarbon dates that span nine ground‐rupturing earthquakes at Pallett Creek. Overall, the AMS dates are similar to and reveal no dramatic bias in the conventional dates. For many layers, however, individual charcoal samples were younger than the conventional dates, leading to earthquake ages that are overall slightly younger than previously reported. New earthquake ages are determined by Bayesian refinement of the layer ages based on stratigraphic ordering and sedimentological constraints. The new chronology is more regular than previously published records in large part due to new samples constraining the age of event R. The closed interval from event C to 1857 has a mean recurrence of 135 years (σ = 83.2 years) and a quasiperiodic coefficient of variation (COV) of 0.61. We show that the new dates and resultant earthquake chronology have a stronger effect on COV than the specific membership of this long series and dating precision improvements from sedimentation rates.