“Alexa learned Arabic”: A translanguaging and multimodal perspective on language and media ideologies

International audience This study offers a novel perspective on translanguaging and multimodality by investigating practices and discourses of a largely unexplored phenomenon of voice assistants in the multifaceted Arab-speaking world. In early 2022, Alexa was launched in "the Khaleeji/Gulf dialect", targeting Saudi Arabian and the United Arab Emirates markets, leading numerous translocal YouTubers to produce videos engaging with the device. Applying the analytical frameworks of multimodal discourse analysis and translanguaging theory, this paper examines an 'unboxing' video, created by an Egyptian YouTuber, and its comment section where Arabic speakers from different Asian and African countries engage in metalinguistic discussions, and position themselves towards Alexa's voice that indexes a specific regional variety of Arabic. They mobilize multimodal and translingual semiotic resources to negotiate the tensions of national, regional, pan-regional and global labels of language and identity. The paper offers critical insights into discourses about Arabic and AI, as mediated on YouTube. We observe a partial reordering of sociolinguistic hierarchies and the emergence of language ideologies tied to global market logics. Voice assistants, as capitalist products, together with their built-in language ideologies, have implications for language perception and sociolinguistic economies.