<p>Rare earth elements (REEs) are critical raw materials due to their essential role in modern technologies. In regolith-hosted REE (RH-REE) deposits, a substantial fraction of the REE is present as ionically adsorbed, exchangeable cations on secondary clay minerals and amenable to mild extraction routes, potentially being less environmentally disruptive than conventional hard-rock REE operations. Shallow drilling in combination with geophysical techniques are used to determine the potential of a regolith resource for economic extraction. In particular, airborne gamma ray spectrometry provides indirect indicators of the regolith and parent lithology by mapping potassium (K), equivalent thorium (eTh) and equivalent uranium (eU) from low-altitude, line-spaced surveys. In a brownfield exploration context, neighboring densely drilled areas can be used as training sets to assess the potential of undrilled areas. To do so, the flightline data need to be interpolated to allow for correlation with REE concentration at borehole locations. In this paper, we show that the traditional methods of deterministic interpolation leads to a biased estimate of the REE concentration. As a solution, we propose a multi-variate stochastic simulation of the radiometric data in a Monte Carlo simulation framework. We illustrate the methodology with a case study from a RH-REE district in south-central Chile, located in the Coastal Range about 15&#xa0;km NE of Concepción. We show how the uncertainty approach is more informative than the deterministic interpolation approach, in terms of identifying promising areas for drilling. Because the large majority of online available geophysical data are deterministically interpolated products, we also conjecture that our findings have wider implications for mineral prospectivity mapping involving geophysical, geological and drilling databases.</p>