Emerging research suggests that caregiving environments and genetic variants independently contribute to social functioning in children with typical development or autism spectrum disorder (ASD). However, biologically plausible interactive models and complimentary assessment of mechanisms are needed to: (1) explain considerable social heterogeneity; (2) resolve inconsistencies in the literature; and (3) develop and select optimal treatments based on individual differences. This study examined the role of child genotypes and responsive parenting in the social development of 104 children with ASD (ages 4-7 years). We utilized a longitudinal, multi-informant design and structural equation models to evaluate: (1) the additive and interactive effects of biologically plausible candidate genes (5-HTTLPR, OXTR, DRD4) and responsive parenting in predicting prospective social development in ASD across three time points spanning 1.5 years; and (2) whether child emotion regulation mediated observed gene x environment interactions (GxEs). Responsive parenting positively predicted prospective change in child social skills; these associations were moderated by 5-HTTLPR and DRD4 in teacher-report models, and DRD4 in parent-report models. No GxE effects were found for OXTR. Emotion regulation did not significantly mediate the GxEs involving 5-HTTLPR and DRD4. Acknowledging the complexities of GxE research, implications for future research, and targeted intervention efforts are discussed. [This is the online version of an article published in "Developmental Psychobiology".]