The role of language in autism spectrum disorder (ASD), more specifically, its function in social communication and strong predictive power on future outcomes, warrants language assessments that have good psychometric properties that capture the heterogeneity of language ability found among diagnosed individuals. Given the rapid growth in intervention and treatment research, there is an urgent need for the development and implementation of outcome measures that are easily obtained and sensitive to change. In this commentary, we argue for the use of natural language samples as measures of expressive language and communication for this purpose and review the literature on their implementation in ASD research. Conceptual and measurement issues are discussed and future developments are outlined.