This study explored cross-situational word learning in children with and without autism spectrum disorder (ASD). Children learn words by mapping object names in caregiver utterances to objects in their visual field. We examined the confluence of caregiver object naming and child visual attention in children with and without ASD at play. Head-mounted eye-tracking revealed that children with ASD spent less time attending to named objects than typically developing (TD) children. In both groups, learning input improved as children accrued increased looking time to named objects across multiple naming events. However, for objects with high quantity of naming events, TD children had higher quality learning input than children with ASD. These findings suggest that the input statistics of social interaction are less conducive to word learning in children with ASD. This work has important implications for clinical interventions to scaffold word learning.