Telepresence robots enhance videoconferencing by enabling the user to ‘move around’ remotely, offering more embodied resources for interaction than speaker gallery-based videoconferencing tools. This study explores interactional troubles related to the affordance of mobility during robot-mediated remote participation in synchronous hybrid classroom interaction. Using multimodal conversation analysis, we analyse video-recorded higher education language teaching classes from Finland in which one student used a Double 3 telepresence robot for remote participation, focusing on how the participants managed activity transitions requiring mediated mobility during one L2 Swedish class. The analysis traces a micro-longitudinal interactional trajectory whereby initial troubles in operating the robot led participants to stop using the remote-controlled mobility feature and resort to carrying the robot from one place to another. We discuss how such a practice of human-assisted mobility emerges interactionally through geographically distributed coordination of the classroom space and what kinds of orientations to the accountability of remote student's action, autonomy and engagement are at stake. The findings illustrate challenges that emerging communication tools such as the telepresence robot can introduce for remote embodied participation and learning, and how teachers and students sometimes come up with unexpected ways of building action with the help of, or despite, technological affordances.