Individuals with Autism Spectrum Disorder (ASD) show a range of language production deficits, however, language comprehension in ASD remains under-studied in part because co-morbid social deficits affect behavioural compliance. This challenge can be overcome by engaging participants in a naturalistic task while passively collecting neural signals. To test predictive processing with naturalistic language, we collect MEG data while 16 8–12-year-old high-functioning participants with a clinical diagnosis of ASD and 16 age- and gender-matched typically developing peers listen to an audiobook story. The neuromagnetic signals are correlated with word-by-word states from a computational model that quantifies incremental sentence predictions in terms of surprisal. Consistent with prior eye-tracking work, our results are compatible with predictive parsing that is equivalent between high-functioning individuals with ASD and TD peers.