<p>Information and communication technology (ICT)-driven platform government has become critical to enhancing bureaucratic responsiveness. Nevertheless, quantitative research examining the impact of artificial intelligence (AI)–powered platform government on screen-level bureaucratic responsiveness remains scarce. Drawing on a dataset of 126,539 public requests from the “government–citizen interaction” portals of Shenzhen’s municipal and district governments (2016–2024), this study investigates the causal effects and temporal dynamics of AI-powered platform government on screen-level bureaucratic responsiveness through an integrated framework that combines regression discontinuity in time (RDiT), the event-study method (ESM), and natural language processing (NLP). The findings indicate that, under enhanced ICT oversight, AI-powered platform government significantly improves screen-level bureaucrats’ response quality. This improvement exhibits a substantial temporal lag and is constrained by organizational inertia, manifesting as a pattern of “lagged short-run efficacy but sustained long-run inefficacy.” Furthermore, while top-down coercive institutional pressures induced by AI-powered platform government do not yield significant gains in overall response efficiency or attitude, they do elevate response quality for consultation requests and at the district level. These results highlight the limitations of relying exclusively on a superior-led, AI-powered platform to catalyze governance improvements. Ultimately, this study underscores the necessity of strengthening the operational capacity of frontline departments handling public requests directly and that of enhancing social self-governance to achieve sustainable administrative responsiveness.</p>