Environmental regulatory pressure, artificial intelligence and corporate green innovation bubbles in China

Environmental regulation aimed at substantive technological transformation often produces visible green innovation activity that grows faster than its quality-anchored complement, creating a green innovation bubble in which patent applications outpace grants. Drawing on regulatory science, decoupling theory, and a reciprocal reading of algorithmic regulation, this study develops a dual-pathway framework in which the same regulatory pressure activates symbolic artificial intelligence deployment that widens the bubble and substantive artificial intelligence deployment that narrows it. Using a panel of 47,185 firm-year observations on Chinese A-share listed firms from 2010 to 2023, with regulatory pressure measured through geocoded counts of nearby environmental supervisory authorities, the analysis confirms that regulatory pressure widens the normalized application-grant gap, with effects robust to instrumental-variable identification, propensity-score matching, and Heckman selection correction. Discursive artificial intelligence intensity positively mediates the relationship while substantive artificial intelligence patenting negatively mediates it. Executive environmental background and regional digital governance attenuate the bubble-amplifying effect, which concentrates among state-owned enterprises, larger firms, and firms outside artificial intelligence pilot zones. The findings reposition the regulation-innovation debate around innovation quality and identify governance levers that bend the dual algorithmic deployment toward substance.