Applications of locative media (e.g., place-based mobile augmented reality [AR]) are used in various educational content areas and have been shown to provide learners with valuable opportunities for investigation-based learning, location-situated social and collaborative interaction, and embodied experience of place (Squire, 2009; Thorne & Hellermann, 2017; Zheng et al., 2018). Mobile locative media applications' value for language learning, however, remains underinvestigated. To address this lacuna, this study employed the widely used construct of language-related episodes (LREs; Swain & Lapkin, 1998) as a unit of analysis to investigate language learning through participation in a mobile AR game. Analysis of videorecorded interactions of four mixed-proficiency groups of game players (two English language learners [ELLs] and one expert speaker of English [ESE] per group) indicates that LREs in this environment were focused on lexical items relevant to the AR tasks and physical locations. Informed by sociocultural theory and conversation analysis, the microgenesis of learners' understanding and subsequent use of certain lexical items are indicated in the findings. This understanding of new lexical items was frequently facilitated by ESEs' assistance and the surrounding physical environment. A strong goal orientation by both ESEs and ELLs was visible, providing implications for task-based language teaching approaches.