In “How AI Sees the City: Urban Visual Intelligence,” published in September 2026 by Routledge, MIT researchers Fábio Duarte, Martina Mazzarello and Carlo Ratti, with Fan Zhang of Peking University, examine how computer vision can turn urban images into data. The book situates the technology within a longer tradition of visual urban study, citing Kevin Lynch and William H. Whyte. Applications discussed include estimating vehicle emissions from traffic-camera footage, traffic flow, intersection safety, street-level activity and greenery.

The authors also flag risks, including intrusive widespread surveillance and the reinforcement of bias through AI systems. They cite roughly 210 cameras per square mile in London and over 5,000 per square mile in Shanghai, and argue safety benefits must be weighed against the erosion of personal freedom. Duarte said AI is not neutral, depending on how it is trained.

Mazzarello said visual tools still require guidance and training. A recent Senseable City study of images from 400,000 Airbnb listings reported that interior design styles retain significant geographic differences rather than becoming globally homogeneous.