Abstract Coupled multi‐decadal global km‐scale simulations completed in recent years open new perspectives in the investigation and understanding of urban climate change. This study introduces a generic method for extracting urban areas and their rural references worldwide and validates it on hourly timescales with remote‐sensing observations of land‐surface temperature for over 400 cities, while identifying the key role of spatial resolution. We showcase the immense amount of hourly data available for each city for 60 years of past and future climate in the Next Generation Earth Modelling Systems simulations, and how these simulations represent surface‐atmosphere interactions driven by urban areas. We finally propose city clustering as a method to exploit these simulations and move beyond city‐specific results. Results suggest a mean potential decrease in future urban heat island intensity driven by stronger rural warming for all clusters except for cold climates.