The shape of nanoparticles can be important in defining their properties. Establishing the exact shape of particles is a challenging task when the particles tend to agglomerate and their size is just a few nanometers. Here we report a structure refinement procedure for establishing the shape of nanoparticles using powder diffraction data. The method utilizes the fundamental formula of Debye coupled with a Monte Carlo-based optimization and has been successfully applied to TiO2-B nanoparticles. Atomistic modeling and molecular dynamics simulations of ensembles of all the ions in the nanoparticle reveal surface hydroxylation as the underlying reason for the established shape and structural features.