Many studies have utilized standardized measures and storybook narratives to characterize language profiles of children with Autism Spectrum Disorder (ASD) and Attention Deficit/Hyperactivity Disorder (ADHD). They report that structural language of these children is on par with mental-age-matched typically developing (TD) peers. Few studies have looked at structural language profiles in conversational contexts. This study examines conversational speech produced in a virtual reality (VR) paradigm to investigate the strengths and weaknesses of structural language abilities of these children. The VR paradigm introduced varying social and cognitive demands across phases. Our results indicate that children from these diagnostic groups produced less complex structural language than TD children. Moreover, language complexity decreased in all groups across phases, suggesting a cross-etiology sensitivity to conversational contexts.