In China, the Pearl River Delta (PRD) plays a leading role as not only an artificial intelligence (AI) innovation hotspot but also a pilot zone for green and low-carbon development. The Super-EBM model was used to measure the PRD’s carbon emission efficiency (CEE) from 2006 to 2021. On this basis, dual fixed effect, mediation effect, and threshold effect regression estimation approaches are used to analyze the influence of AI on CEE and its internal mechanism. The results show that AI can significantly improve the CEE, and this conclusion remains true after endogenous and robustness tests such as difference-in-difference (DID), time lag effect, independent variable replacement, and split‑sample tests. Mechanism analysis reveals that industrial structure upgrading and energy efficiency are two basic paths for improving CEE. The analysis of the panel threshold regression model and heterogeneity test shows that with industrial structure upgrading and energy efficiency improvement, AI has a more significant effect on promoting CEE, with that effect being more prominent in the PRD’s core cities. The government should vigorously promote the deep integration of AI and the low-carbon economy, give full play to the indirect driving role of industrial structure upgrading and energy efficiency, strengthen regional cooperation, promote the coordinated development of various regions, and implement differentiated low-carbon transformation policies.