Artificial Intelligence and Corporate ESG Performance: Causal Evidence from Machine Learning Methods

This study examines the effect of artificial intelligence (AI) innovation on corporate ESG performance in Chinese listed firms. Using 42,996 firm-year observations of 4792 A-share listed firms from 2009 to 2023, we measure AI innovation by AI patent applications and ESG performance by Huazheng ESG ratings. Two-way fixed effects estimates indicate that a one-unit increase in ln(1 + AI patent applications) is associated with a 0.037-point increase in the ESG score, or about 5% of a standard deviation for a one-standard-deviation increase in AI. Double/Debiased Machine Learning (DML) estimates with firm-level cross-fitting are similar (0.026–0.038). The relationship is robust to lagged AI, predetermined controls, ordinal and rank-based outcomes, and an event-study design that shows no differential ESG trends before firms’ first AI patent application; however, it is driven by the intensity of AI patenting rather than by the extensive margin and is imprecisely estimated within the pre- and post-2017 disclosure regimes. Mechanism tests suggest efficiency and internal-control channels, whereas the management expense ratio does not mediate the effect. The effect is stronger in larger firms and in firms with greater green innovation capability and tends to be larger in more market-oriented regions (significant at the 10% level), while the difference between heavily polluting and other firms is not statistically significant. Under a conditional exogeneity assumption, these results are consistent with a causal effect of AI innovation on ESG performance.