Abstract: The calculation of defect levels in semiconductors within a density functional theory approach suffers greatly from the band gap problem. We propose a band gap correction scheme that is based on the separation of energy differences in electron addition and relaxation energies. We show that it can predict defect levels with a reasonable accuracy, particularly in the case of defects with conduction band character, and yet is simple and computationally economical. We apply this method to ZnO doped with group III elements (Al, Ga, In). As expected from experiment, the results indicate that Zn substitutional doping is preferred over interstitial doping in Al, Ga, and In-doped ZnO, under both zinc-rich and oxygen-rich conditions. Further, all three dopants act as shallow donors, with the +1 charge state having the most advantageous formation energy. Also, doping effects on the electronic structure of ZnO are sufficiently mild so as to affect little the fundamental band gap and lowest conduction bands dispersion, which secures their n-type transparent conducting behavior. A comparison with the extrapolation method based on LDA+U calculations and with the HeydScuseriaErnzerhof hybrid functional (HSE) shows the reliability of the proposed scheme in predicting the thermodynamic transition levels in shallow donor systems.