Where does attention go when AI listens? Meta-awareness of mediated attention

Ambient artificial intelligence (AI) can translate speech, generate notes, and summarize an event while it unfolds. These systems may improve access and task performance without leaving the user’s experience unchanged. This article develops meta-awareness of mediated attention as a construct for explaining whether users recognize that an AI-generated representation is shaping what they attend to and what becomes available to experience. The underlying process is experiential reconfiguration: attention is redistributed between an ongoing event and the representation produced from it. Reconfiguration may scaffold contact when assistance makes otherwise inaccessible information available, support concurrent augmentation when event and representation remain jointly relevant, or produce representational substitution when the generated output becomes the primary attentional object. Each configuration is judged relative to the user’s unassisted access. Evidence from photo-taking, captioning, translation, and AI-supported note-taking is consistent with selective redistribution rather than uniform cognitive impairment, although direct evidence concerning ambient generative AI remains limited. The proposed model separates mediation, attentional allocation, meta-awareness, regulation, and outcomes. Meta-awareness creates the possibility of deliberate regulation but does not guarantee that users will redirect attention. The article advances propositions concerning representational orientation, conditional scaffolding and substitution, dissociations between semantic and contextual outcomes, awareness calibration, and the conditions under which awareness supports regulation. Design should support recoverable contact with the event. The framework shifts evaluation away from whether AI increases or decreases attention in the abstract and toward how users attend through AI, what the representation privileges or omits, and whether this mediation remains recognizable and governable.