Abstract: Disruptive green innovation (DGI) is essential for achieving carbon neutrality goals; however, its development is constrained by the dual challenges of externalities and substantial knowledge burdens. Artificial Intelligence (AI), with its potential to act as an “invention of a method of invention,” raises important yet underexplored questions concerning whether and how it influences firms’ DGI. Drawing on the knowledge recombination perspective, this study develops a theoretical framework to elucidate the mechanisms through which AI technologies can facilitate DGI at the firm level. To empirically validate this framework, we utilize panel data from China’s A-share listed manufacturing firms spanning 2007 to 2019. To improve measurement accuracy, we employ the BERT machine learning model to analyze corporate annual reports and construct firm-level AI adoption indicators, while the disruptiveness of green innovation is measured using the CD index derived from patent citation networks. The results reveal that AI significantly promotes DGI, primarily through dual pathways of expanding knowledge search breadth and deepening knowledge search depth. Heterogeneity analysis demonstrates that AI’s positive effects on DGI are more pronounced in regions with advanced digital infrastructure, weaker environmental regulation intensity, and firms receiving higher R&D subsidies. This study deepens the understanding of AI’s role in driving DGI at the micro-enterprise level, providing empirical evidence and theoretical insights for fostering synergistic development of intelligent and green transformation in corporations.

Key words: artificial intelligence, disruptive green innovation, knowledge recombination, machine learning