<p>This article argues that machine perception should be understood not merely as a technical development in artificial intelligence, but as a cultural technique that reorganizes the societal conditions of visibility. In AI-mediated exhibitionary environments, vision no longer functions primarily as representation for human spectators; it becomes an infrastructural operation oriented toward alignment, persistence, and inference. Drawing on examples from augmented museum systems and immersive installations, the article introduces the concept of infrastructural visibility to describe how algorithmic thresholds delimit what can appear, persist, and be recognized within shared environments. Rather than focusing solely on surveillance or control, the argument reframes machine seeing as a transformation in the governance of perception itself. As computational systems increasingly mediate recognition and participation, they redistribute perceptual labor, reshape epistemic authority, and reorganize the normative foundations of shared visibility. The societal challenge is therefore not only how AI systems classify the world, but how perceptual infrastructures become authoritative and how they might be rendered accountable.</p>