There is little research on the production of speech disfluencies such as silent pauses, repetitions, self-corrections, and filled pauses (e.g., eh, em) in monolingual autistic children, and there is no data on this crucial part of speech production in bilingual autistic children. This study aims to address this gap by examining disfluency production in bilingual autistic and non-autistic children across two linguistically distinct languages, HL-Russian (the home language) and SL-Hebrew (the societal language). Fifty-one bilingual Russian-Hebrew-speaking autistic and non-autistic children aged 5–9 (autistic: n = 21; non-autistic: n = 30), matched for age and non-verbal intelligence, participated in picture-based story-generation tasks (LITMUS MAIN, Gagarina et al., ZAS Papers in Linguistics, 63:1–36, 2019). Audio recordings of narrative samples were transcribed, coded, and scored for eleven disfluency types using CLAN tools. The non-autistic group produced higher overall disfluency rate than the autistic group. The autistic group exhibited fewer filled and silent pauses than the non-autistic group in HL-Russian. Furthermore, non-autistic children manifested varied distribution of disfluency types across languages, while autistic children displayed more consistent patterns across languages. In summary, we replicated findings from previous research on monolinguals only partly, as no between-group difference in filled pauses was found in SL-Hebrew. Additionally, bilingual autistic children exhibited language-universal patterns of disfluency production, whereas their non-autistic peers displayed language-specific patterns.