Firms using AI see stronger environmental and governance performance

Artificial intelligence (AI) adoption is associated with stronger corporate environmental, social and governance (ESG) performance, with the sharpest gains seen in environmental and governance outcomes, reveals new research published in Administrative Sciences. 

The study, titled “How Does Artificial Intelligence Improve Corporate ESG Performance?”, examines 1,774 Chinese A-share listed companies from 2012 to 2023, using annual report disclosures, ESG ratings, patent data and financial indicators to assess whether AI transformation changes corporate sustainability performance.

AI adoption shows measurable ESG gains

The researchers found that firms undergoing AI transformation recorded higher ESG performance than companies that didn't show evidence of substantive AI deployment. The analysis identified AI transformation through annual report disclosures, but excluded vague references, broad technology outlooks and symbolic statements that did not point to real business application.

Using a difference-in-differences framework, the study estimated that AI transformation was linked to an average 0.206-point increase in ESG scores after controlling for firm characteristics, company-level fixed effects and annual shocks. The finding remained stable across several tests, including alternative AI measures, alternative ESG ratings, placebo tests, propensity score matching, additional fixed effects and methods designed to address treatment-effect variation across firms and years.

The impact was not evenly spread across ESG dimensions. AI transformation had a statistically significant positive association with environmental performance and governance performance, but its effect on social performance was not significant. That pattern suggests that AI’s current corporate sustainability role is concentrated more in resource efficiency, emissions-related management, monitoring, internal controls and transparency than in wider social responsibility areas such as labor relations, external stakeholder engagement or community impact.

The authors treated AI transformation not as a narrow technology purchase, but as a broader organizational change involving data, algorithms, computing capacity and business processes. Firms may mention AI in public documents without changing operations. To reduce that risk, the study checked whether text-based AI disclosure was supported by observable AI investment intensity. When stricter definitions were used, combining disclosure with AI-related capitalized investment, the positive ESG relationship still held.

Green innovation and governance emerge as key channels

The research identifies two main pathways through which AI transformation may lift ESG performance: green innovation and governance improvement.

Green innovation

AI appears to support green innovation by improving firms’ ability to process information, detect inefficiencies, optimize resource use and speed up technology search. The study used green utility patents as a measure of green innovation and found that AI-transformed firms generated higher green patent output. When green innovation was added to the ESG model, it remained a significant positive factor, supporting the view that AI strengthens ESG performance partly by helping firms develop cleaner processes and technologies.

This is crucial for companies facing pressure to move from symbolic sustainability pledges to measurable environmental performance. AI tools can help identify high-energy-use operations, track emissions-related processes, improve production decisions and support green research and development. The research indicates that these operational gains can translate into broader ESG improvements.

Governance improvement

The governance pathway was measured through stock price crash risk, which the study used as a proxy for governance problems tied to opacity, bad-news hoarding and weak monitoring. AI transformation was associated with lower crash risk, indicating that firms adopting AI may improve internal monitoring, risk alerts, data integration and transparency. Lower crash risk was also linked to stronger ESG performance, suggesting that AI can support sustainability not only through cleaner operations, but also by improving the information environment inside firms.

However, the governance channel was weaker than the green innovation channel. The authors found evidence that AI reduces crash risk and that lower crash risk is associated with higher ESG performance, but the indirect governance effect was only marginally significant in some tests. The result still points to a governance benefit, but one that should be read with more caution than the environmental mechanism.

The study also addressed concerns over causality and AI washing. It used China’s National AI Innovation and Development Pilot Zone policy as part of an instrumental-variable strategy, linking regional AI policy exposure with firms’ earlier digital foundations. The results continued to point to a positive AI-ESG relationship, though the authors cautioned that policy shocks may influence ESG through other channels as well, including regulatory scrutiny or incentives.

Stronger impact seen in state-owned, lower-tech and heavily polluting firms

The ESG gains from AI transformation were strongest among state-owned enterprises, non-high-tech firms and heavily polluting companies. For state-owned enterprises, the effect was larger and statistically significant, while it was not significant for non-state-owned firms. The result suggests that companies with closer links to policy priorities and stronger institutional responsibilities may use AI more directly for environmental management, compliance and governance improvement. State-owned firms may also face stronger pressure to align business transformation with public sustainability goals.

The research also found a stronger AI-ESG effect among non-high-tech firms than high-tech firms. The explanation lies in the size of the marginal gain. High-tech firms may already have strong digital systems, advanced management processes and higher innovation capacity, leaving less room for AI to create immediate ESG improvements. Non-high-tech firms, on the other hand, may gain more from AI because it helps modernize older production systems, improve resource allocation and close digital gaps.

Heavily polluting firms also showed a stronger response. These companies face greater regulatory, reputational and operational pressure to improve environmental performance. AI adoption may help them monitor emissions, improve energy use, support pollution-control decisions and meet compliance demands more effectively. In firms with higher environmental risk, the sustainability value of AI may therefore be more visible and urgent.

The study suggests that AI’s ESG value depends on ownership structure, industry position and environmental pressure. The same technology may produce stronger sustainability gains where firms face clearer regulatory incentives, larger operational inefficiencies or greater public scrutiny.

The authors acknowledge that annual report disclosures may not fully capture the depth of AI transformation, even after efforts to screen out empty AI language. They also noted that the evidence should not be treated as absolute proof of causality, despite the use of multiple identification and robustness methods. Future research could strengthen measurement by using more direct indicators such as AI patents, workforce composition, IT infrastructure and detailed investment data.