Background: Language abilities are highly heterogeneous in autism. While a multimodal assessment approach is recommended to capture the language variability, this is not always possible. Therefore, it is important to gain contextual information about different language assessments to determine which assessment is most appropriate for different research questions. As most current work is based on English-speaking populations, this paper compares three language assessment modalities (standardized assessment, parent survey, and a natural language sample) between English-speaking and Dutch-speaking autistic and neurotypical children. Method: The Mullen Scales of Early Learning, the Vineland Adaptive Behavior Scales and a naturalistic language sample were employed to measure language in 100 preschool-aged participants. Correlation analyses and mixed-effect regressions were conducted, and Bland-Altman plots were created to examine the similarity between measures. Results: English-speaking and Dutch-speaking parents rated their children's expressive language higher than their receptive language. The best agreement between measures was for standardized language and parent report. Agreement was higher for children with low language scores. Primary language (English vs. Dutch) did not significantly affect the results, but age, nonverbal cognitive abilities and biological sex were shown to predict expressive and receptive language outcome. Conclusions: For autistic children with low language levels, parent-reported and standardized language measures provide researchers with similar information. Depending on the available time and resources, researchers may choose to use one of these methods. However, for autistic children with (above) average language abilities, multiple modalities should be considered to gain a comprehensive understanding of their language abilities across different settings. A natural lan- guage sample is of most added value next to a standardized assessment or parent report.