Phototoxicity is of increasing concern in dermatology, since modern lifestyle is often associated with exposure to sunlight. The most commonly reported process is via oxidative reactions. Therefore characterizing the “photo-pro-oxidant” potential of a compound early in its industrial development is of utmost interest, especially for compounds likely to undergo sunlight exposure in skin. Today there is a need for filtering compounds to be tested in the 3T3 neutral red uptake in vitro test for phototoxicity since testing requires resources. A computational model aiming at predicting the mechanisms that imply the generation of reactive oxygen species was developed using a diverse set of 56 chemicals having 3T3 NRU data. An historical mechanistic (Q)SAR model developed for polycyclic aromatic hydrocarbons was used to derive the new mechanistic model: descriptors were selected upfront to describe the modeled phenomenon. The historical parabolic relationships between phototoxicity and the energy gap ( E GAP) between energies of the highest occupied molecular orbital and the lowest unoccupied molecular orbital was confirmed. The model predicts chemicals to be “phototoxic or photodegradable”, or “non-phototoxic and non-photodegradable”. A four-step testing strategy is proposed to enable the reduction of experimental testing with the in silico model implemented as a first screen.