Instrumented mouthguard systems are used to study rigid body head kinematics in vivo in a variety of athletic environments. Previous work has assessed these systems when rigidly attached to anthropomorphic test device (ATD) heads and found good fidelity compared with reference sensors. Here, we examine the fidelity of two instrumented boil-and-bite mouthguards (iMGs) in helmeted, human cadaver impact tests at three velocities and three impact locations. The Prevent Biometrics iMG correlates with reference linear acceleration up to approximately 60 g, underestimating impacts at higher magnitudes. The Prevent iMG correlates well with angular velocity, while underestimating angular acceleration past approximately 3,000 rad/s². The DTS iMG consistently overestimates angular acceleration and velocity, and transformed linear kinematics are not well-correlated to reference kinematics at the head centre of gravity (HCG) for these impact conditions. Future work should examine the influence of the mandible and subject-specific mouthguard fit on fidelity of impact kinematics. This study highlights the need for cadaver validation studies of wearable head impact instrumentation systems, particularly when considering realistic (non-idealised) sensor-skull coupling and accounting for interactions with the mandible.