Instead, it shows that AI urbanism is a contested political and discursive formation 鈥 a set of truth-claims about what AI is, what it can do, and why it should govern. Using a Foucauldian discourse analysis, the paper asks how ideas such as autonomy become credible and actionable in urban governance, and how they reshape who gets to decide, what counts as rational policy, and whose futures are made governable in AI鈥檚 name. In this way, the paper shifts focus away from technical capability alone and towards the politics of how AI is made legitimate.
A major contribution of the paper is to show that AI urbanism does not arrive on a blank slate. It grows out of what the authors call the 鈥渟martmentality鈥 of the smart city, but it also fragments into distinct urban AI formations rather than following one universal model. Through a comparative analysis of Shenzhen, Boston and Barcelona, the paper identifies three contrasting formations: Shenzhen鈥檚 predictive orchestration, Boston鈥檚 careful and modest counter-conduct, and Barcelona鈥檚 procedurally bounded conditionality. These are not treated as local variations of the same template, but as divergent projects that organise authority, accountability, participation and temporality in different ways.
The paper鈥檚 distinctive theoretical intervention is therefore to pluralise 鈥淎I urbanism鈥 into 鈥淎I urbanismS鈥. It introduces an abductively developed six-facet analytical framework that helps scholars examine how AI鈥檚 truth-claims are assembled, authorised and contested across different urban settings. In doing so, the paper not only extends Foucauldian debates on regimes of truth and governmentality into the study of urban AI, but also provincialises dominant narratives of the 鈥渁utonomous city鈥. Its broader message is that there is nothing inevitable about current AI futures in cities; rather, they are historically situated, politically constructed, and therefore open to challenge, reworking, and alternative imaginaries.
A major contribution of the paper is to show that AI urbanism does not arrive on a blank slate. It grows out of what the authors call the 鈥渟martmentality鈥 of the smart city, but it also fragments into distinct urban AI formations rather than following one universal model. Through a comparative analysis of Shenzhen, Boston and Barcelona, the paper identifies three contrasting formations: Shenzhen鈥檚 predictive orchestration, Boston鈥檚 careful and modest counter-conduct, and Barcelona鈥檚 procedurally bounded conditionality. These are not treated as local variations of the same template, but as divergent projects that organise authority, accountability, participation and temporality in different ways.
The paper鈥檚 distinctive theoretical intervention is therefore to pluralise 鈥淎I urbanism鈥 into 鈥淎I urbanismS鈥. It introduces an abductively developed six-facet analytical framework that helps scholars examine how AI鈥檚 truth-claims are assembled, authorised and contested across different urban settings. In doing so, the paper not only extends Foucauldian debates on regimes of truth and governmentality into the study of urban AI, but also provincialises dominant narratives of the 鈥渁utonomous city鈥. Its broader message is that there is nothing inevitable about current AI futures in cities; rather, they are historically situated, politically constructed, and therefore open to challenge, reworking, and alternative imaginaries.