Deconstructing the green productivity paradox: How AI-enabled green innovation shapes efficiency and technological change

Manufacturing firms in China have become more active in AI-enabled green innovation (AIGI) with the support of the dual‐carbon agenda and increasingly strict environmental regulation. Yet the productivity returns to AIGI remain uneven. This study examines whether AIGI can explain differences in green total factor productivity (GTFP) and contribute to narrowing such gaps. Using firm-level data from Chinese manufacturing firms and double machine learning, the results reveal an asymmetric pattern: AIGI improves GTFP mainly through green efficiency change (GEC), while its short-run association with green technological change (GTC) is negative. This suggests that early productivity gains arise less from frontier-level green breakthroughs than from embedding AIGI into production, management, and organizational routines. The effect is more pronounced in pollution-intensive industries, state-owned enterprises, and firms outside eastern China. Mechanism tests confirm that the efficiency channel operates through improved resource allocation and a shift of innovation effort toward application and deployment. Digital strategy intensity weakens these gains and further strengthens the negative effect on GTC, reflecting integration burdens and capability rigidity in highly digitized firms.