Jun Zhang, au J. Cox, Jing Wang
Urban artificial intelligence (AI) is often cast as an inevitable upgrade to smart cities; this paper instead treats it as a contested regime of truth that reorders how cities are governed. Using Foucauldian discourse analysis, we trace how autonomy is assembled and authorized as a truth-claim in three sites: Shenzhen, Boston and Barcelona. We show how AI urbanism grows out of smart city smartmentality but fragments into what we call distinct “urban AI formations’ – Shenzhen’s predictive orchestration, Boston’s modest counter-conduct, and Barcelona’s procedurally bounded conditionality. Building an abductively derived, six-facet analytical framework, we argue that these formations are not local variants of a single model but divergent projects that redefine what counts as rational governance, and for whom. The paper thus pluralizes AI urbanism into AI urbanismS, provincializing dominant narratives of “autonomous cities’ and offering a transferable heuristic for examining how AI’s truth-claims become credible, governable and open to contestation in different urban contexts.