Corporate Artificial Intelligence and Green Innovation Efficiency: The Roles of Environmental Disclosure, R&D Human Capital and Government Subsidies

Artificial intelligence (AI) is widely promoted as a tool for corporate green transformation, but whether it helps firms convert research and development (R&D) resources into green technologies more efficiently remains unclear. Drawing on the resource-based view and information-asymmetry theory, this study examines whether corporate AI is associated with green innovation efficiency (GEff), whether environmental information disclosure quality (Eidq) transmits this association, and whether government subsidies condition it. Using 28,219 firm-year observations of Chinese A-share listed firms from 2012 to 2024, we measure AI with a text-based index from annual reports and estimate two-way fixed-effects models, complemented by leave-one-out instrumental variables, dynamic specifications, alternative measures including AI patents and bootstrap mediation tests. AI is positively and robustly associated with GEff, an association driven mainly by green invention patents, whereas past GEff does not predict future AI. Eidq and R&D human capital provide significant but modest partial mediating channels, and the AI–GEff association is stronger at higher subsidy levels, particularly for non-innovation subsidies. Formal tests reveal no significant differences across regions, pollution intensity, technology status, life-cycle stages, industries or ownership. The findings suggest that combining AI deployment with credible environmental disclosure and subsidies for AI adoption can support more efficient green innovation.