The accuracy of a novel approach for evaluating induction energies is probed. Fully distributed polarizability models obtained within the theory of atoms in molecules (AIM) of Bader are used to generate large grids of induction energies much more rapidly than with a finite perturbation approach. Next, compact models of distributed polarizabilities are fitted to reproduce the target induction energies. Test calculations show that this method, based on one second-order Moller-Plesset (MP2) calculation and the subsequent determination of fully distributed polarizabilities, is able to provide compact, yet optimal distributed polarizability models.