Skeletal muscles have inertia that leads to inertial forces acting around joints. Although these inertial muscle forces contribute to joint kinetics, they are not typically accounted for in musculoskeletal models used for human movement biomechanics research. Ignoring inertial forces can lead to errors in joint kinetics, but how large these errors are in inverse dynamics calculations of common movements is yet unclear. We, therefore, examined the role of shank muscle inertia on ankle joint moments during the swing phase of running at different speeds. A custom musculoskeletal modelling and simulation platform was used to perform inverse dynamics with a model that either combined muscle mass in the total shank mass, or considered the gastrocnemius lateralis/medialis, soleus, and tibialis anterior muscles as separate masses from the shank. Ankle moments were considerably affected when muscles were modelled as separate masses, with a general shift towards reduced dorsiflexion and higher plantarflexion moments. Differences between both modelling conditions increased with running speed and ranged between 0.8 and 1.6 Nm (ankle moment profile root mean square error), 8–18 % (peak dorsiflexion moment difference) and 24–42 % (peak plantarflexion moment difference). Moreover, we observed a complex combination of inertial forces, especially those due to rotation and translation of the shank, in which the direction of inertial force changed during the swing phase. These results show that ignoring muscle inertia in musculoskeletal models can lead to under- or overestimations of structure-specific loads and thus erroneous study conclusions. Our results suggest that muscle inertial forces should be carefully considered when using musculoskeletal models.