Head impact sensors worn in the mouth are popular because they couple directly to the teeth and provide six-degree-of-freedom head measurements. Mouthpiece signal filters have conventionally used cutoff frequencies lower than recommended practices (Society of Automotive Engineers, SAE J211-1) to eliminate extraneous noise when measuring with live subjects. However, there is little information about the effects of filter choice on the accuracy of signals measured by instrumented mouthpieces. Lack of standardization in head impact measurement device post-processing techniques can result in data that are not comparable across studies or device brands. This study sought optimal filter cutoff frequencies for six-degree-of-freedom measurements made at the teeth using instrumented mouthguards. We collected linear acceleration and angular velocity signals at the head center of gravity (CG) using laboratory-grade instrumentation. We also collected and filtered similar six-degree-of-freedom measurements from an instrumented mouthguard using 24 cutoff frequencies, from 25 to 600 Hz. We transformed the measurements to linear acceleration at the center of gravity of the head (CG) using all kinematic variables at the teeth, optimizing linear and angular mouthguard cutoff frequencies with one equation. We calculated the percent error in transformed peak resultant linear acceleration and minimized the mean and standard deviation in error. The optimal cutoff frequencies were 175 Hz for linear acceleration and 250 Hz for angular velocity. Rigid impacts (3–5 ms duration) had higher optimal cutoff frequencies (175 Hz linear acceleration, 275 Hz angular velocity) than padded impacts (10–12 ms duration; 100 Hz linear acceleration, 175 Hz angular velocity), and all impacts together (3–12 ms duration; 175 Hz linear acceleration, 250 Hz angular velocity). Instrumented mouthpiece manufacturers and researchers using these devices should consider these optimal filter cutoff frequencies to minimize measurement error. Sport-specific filter criteria for teeth-based sensors may be warranted to account for the difference in optimal cutoff frequency combination by impact duration.