The theoretical description of the electronic structure of magnetic insulators and, in particular, of transition-metal oxides (TMOs), MnO, FeO, CoO, NiO, and CuO, poses several problems due to their highly correlated nature. Particularly challenging is the determination of the band gap. The most widely used approach is based on density functional theory (DFT) Kohn-Sham energy levels using self-interaction-corrected functionals (such as hybrid functionals). Here, we present a different approach based on the assumption that the band gap in some TMOs can have a partial Mott-Hubbard character and can be defined as the energy associated with the process Mm+(3dn) + Mm+(3dn) → M(m+1)+(3dn-1) + M(m-1)+(3dn+1). The band gap is thus associated with the removal (ionization potential, I) and addition (electron affinity, A) of one electron to an ion of the lattice. In fact, due to the hybridization of metal with ligand orbitals, these energy contributions are not purely atomic in nature. I and A can be computed accurately using the charge transition level (CTL) scheme. This procedure is based on the calculation of energy levels of charged states and goes beyond the approximations inherent to the Kohn-Sham (KS) approach. The novel and relevant aspect of this work is the extension of CTLs from the domain of point defects to a bulk property such as the band gap. The results show that the calculation based on CTLs provides band gaps in better agreement with experiments than the KS approach, with direct insight into the nature of the gap in these complex systems.