It remains difficult to obtain reliable geochemical signatures of uncontaminated kimberlite magma from bulk rock studies due to the combined effects of crustal assimilation and element mobility during post-emplacement alteration processes. Groundmass perovskite (CaTiO3), a typical accessory phase, from Orapa (Botswana) and Wesselton (South Africa) kimberlites has been used to evaluate the isotope and trace element composition of the pre-contamination magmas and the effects of shallow level contamination. In-situ trace element signatures of Orapa and Wesselton perovskite grains are broadly similar and unaffected by crustal contamination. Single grain 87Sr/86Sr isotope ratios of perovskite from Orapa (0.7030–0.7036) are less scattered than bulk rock analyses (0.7063–0.7156), which are variably affected by contamination and late stage alteration. Initial 87Sr/86Sr isotope ratios of perovskite (0.7044–0.7049) from Wesselton overlap with published whole rock studies on fresh hypabyssal kimberlites (0.7042–0.7047). The limited intra-kimberlite variation in Sr isotope ratios recorded by the perovskite are unlikely to be due to crustal contamination as the calculated liquid compositions in equilibrium with the perovskite analysed typically have > 1500 ppm Sr, and most common crustal lithologies underlying these kimberlites have relatively low Sr contents and are not highly radiogenic. Calculated pre-shallow level contamination magma compositions for Orapa and Wesselton have significantly fractionated LREE and highly variable non-smooth trace element patterns. Initial Sr and Nd isotope ratios of both kimberlites fall on the mantle Nd–Sr array with enriched Sr and slightly depleted Nd signatures, similar to Group I kimberlites. Overall, the trace element and isotopic composition of Orapa and Wesselton kimberlites are similar to the reported Group I kimberlites from southern Africa, which are derived by very low degrees of partial melting from a LREE depleted metasomatised sub-continental lithospheric mantle (SCLM) source.