Assuring the safety of personal care products without testing in animals is a goal common to many personal care products manufacturers, due to both ethical concerns for animal welfare, as well as the limited relevancy of animal models to predict human responses. Towards this goal, the cosmetics and personal care industry has increasingly relied upon human cell-based 3–dimensional reconstructed tissues to evaluate the safety of their product candidates in various target tissues. Accordingly, we have developed a program for screening the potential irritancy of teeth whitening products in oral mucosal tissues using commercially-available oral buccal (SkinEthic RHO) and gingival (SkinEthic RHG) models. Four formulations containing H2O2 at various concentrations and one sodium bicarbonate were tested at four exposure times to determine ET50 values. Three irritancy endpoints were measured after each exposure: viability using the MTT conversion assay, the amount of IL-1a released from the tissues, and histological changes. Both models presented the same rank order of the five materials, with increases in the ET50 values correlating with decreases in H2O2 concentration. The formula containing sodium bicarbonate, however, was non-toxic in both models. Histological analysis confirmed the MTT results and provided evidence of the chemical impact upon cellular and tissue morphology. IL-1a release did not appear to be as sensitive as the MTT assay at shorter exposure times, although it may be useful to differentiate among formulations predicted by the MTT assay to be of low irritation potential. Our results suggest that the MTT viability and histology endpoints in 3–D human oral reconstructed tissues can provide useful predictive information to support an oral care products safety program.