Inherited retinal dystrophies (IRDs) comprise a large group of genetically and clinically heterogeneous diseases that lead to progressive vision loss, for which a paucity of disease-mimicking animal models renders preclinical studies difficult. We sought to develop pertinent human cellular IRD models, beginning with choroideremia, caused by mutations in the CHM gene encoding Rab escort protein 1 (REP1). We reprogrammed REP1-deficient fibroblasts from a CHM (-/y) patient into induced pluripotent stem cells (iPSCs), which we differentiated into retinal pigment epithelium (RPE). This iPSC-derived RPE is a polarized monolayer with a classic morphology, expresses characteristic markers, is functional for fluid transport and phagocytosis, and mimics the biochemical phenotype of patients. We assayed a panel of adeno-associated virus (AAV) vector serotypes and showed that AAV2/5 is the most efficient at transducing the iPSC-derived RPE and that CHM gene transfer normalizes the biochemical phenotype. The high, and unmatched, in vitro transduction efficiency is likely aided by phagocytosis and mimics the scenario that an AAV vector encounters in vivo in the subretinal space. We demonstrate the superiority of AAV2/5 in the human RPE and address the potential of patient iPSC-derived RPE to provide a proof-of-concept model for gene replacement in the absence of an appropriate animal model.