The impact of artificial intelligence applications on improving energy efficiency and avoiding energy rebound: Evidence from China's “national smart city pilot” policy

Artificial intelligence (AI) applications in China's smart city initiatives are widely regarded as having the potential to enhance energy efficiency, yet whether such applications may trigger or exacerbate the energy rebound effect remains an open empirical question. Using panel data for 279 Chinese prefecture-level cities, this paper examines the impact of AI applications on urban energy efficiency under the Smart City Pilot policy by employing a staggered DID approach. The results show that AI applications promoted by smart city pilots significantly reduce urban energy intensity and improve energy efficiency, and these findings remain robust across a series of robustness checks. Further heterogeneity analysis reveals that the energy-efficiency-enhancing effects of AI are significantly stronger in northern cities, non–old industrial base cities, non-resource-based cities, and cities that are not designated as key environmental protection areas.In terms of underlying mechanisms, this study takes electricity as an illustrative case and finds that AI applications significantly improve electricity use efficiency without leading to a statistically significant increase in electricity consumption or total energy consumption. These results suggest that, during the sample period, the energy efficiency gains associated with AI applications are not offset by additional energy use, and no statistically detectable energy rebound effect is observed. Overall, this paper provides empirical evidence on how AI applications in smart city development can enhance urban energy efficiency and, to some extent, mitigate the energy rebound effect, offering useful insights for related policy design. • Employ a staggered DID framework to identify the policy effect of AI applications. • Utilize panel data from 279 Chinese prefecture-level cities to ensure robust empirical inference. • AI applications significantly reduce energy intensity and improve urban energy efficiency. • No statistically significant evidence of energy rebound is observed during the sample period. • The findings provide policy-relevant insights for energy transitions in developing economies.