In the context of Social Stories™, we compared visual attention to social scenes using BoardMaker™ versus photographic stimuli among typically developing (TD) children and age-matched children with ASD. For visual attention, the dependent measures were the number of fixations and fixation time to eye, mouth, and 'other' (background) areas of interest. We found no differences between groups when viewing faces alone but when viewing complex scenes (i.e., people and objects) an interaction was observed: the TD and ASD groups were no different in the BoardMaker™ condition but were different in the photograph condition. Specifically, we found 1) a tendency toward more mouth-looking in the photograph condition among children with ASD, which was negatively correlated with attention shifting and verbal IQ; and 2) a tendency toward more 'other'-looking (i.e., background regions), which was negatively correlated with attention shifting, age, and central coherence. We argue that atypical scanning of complex scenes in ASD may reflect resource allocation strategies: as the complexity of image content increases, the simpler and more efficient the scan strategy.