SUMMARY The Earth's mantle transition zone (MTZ) plays a key role in the thermal and compositional interactions between the upper and lower mantle. Seismic anisotropy provides useful information about mantle deformation and dynamics across the MTZ. However, seismic anisotropy in the MTZ is difficult to constrain from surface wave or shear wave splitting measurements. Here, we investigate the sensitivity to anisotropy of a body wave method, SS precursors, through 3-D synthetic modelling and apply it to real data. Our study shows that the SS precursors can distinguish the anisotropy originating from three depths: shallow upper mantle (80–220 km), deep upper mantle above 410 km, and MTZ (410–660 km). Synthetic resolution tests indicate that SS precursors can resolve $\ge $3 per cent azimuthal anisotropy where data have an average signal-to-noise ratio (SNR = 7) and sufficient azimuthal coverage. To investigate regional sensitivity, we apply the stacking and inversion methods to two densely sampled areas: the Japan subduction zone and a central Pacific region around the Hawaiian hotspot. We find evidence for significant VS anisotropy (15.3 ± 9.2 per cent) with a trench-perpendicular fast direction (93° ± 5°) in the MTZ near the Japan subduction zone. We attribute the azimuthal anisotropy to the grain-scale shape-preferred orientation of basaltic materials induced by the shear deformation within the subducting slab beneath NE China. In the central Pacific study region, there is a non-detection of MTZ anisotropy, although modelling suggests the data coverage should allow us to resolve at least 3 per cent anisotropy. Therefore, the Hawaiian mantle plume has not produced detectable azimuthal anisotropy in the MTZ.