The binding of N-heterocyclic carbenes to Ce(III) and U(III) compounds is characterized by quantum chemical methods. Density functional methods are in qualitative agreement with experiment that binding to U(III) is more favorable than to Ce(III); after correcting for basis-set superposition error, quantitative agreement with experiment is achieved with a multireference second-order perturbation theory approach accounting for relativistic effects. The small computed (and observed) preference derives from a combination of several small effects, including differences in electronic binding energies, rovibrational partition functions, and solvation free energies. Prospects for ligand modification to improve the differentiation between lanthanides and actinides are discussed on the basis of computational predictions.