Artificial Intelligence and Corporate ESG Performance: A Mechanism Analysis Based on Corporate Efficiency and External Environment

Abstract

The rapid advancement of artificial intelligence (AI) has become a key driver in shaping firms’ environmental, social, and governance (ESG) performance. This study investigates the impact of corporate AI capabilities on ESG outcomes and examines how external environmental factors moderate this relationship. Using panel data from all A-share listed firms on the Shanghai and Shenzhen Stock Exchanges between 2010 and 2023, we measure firms’ AI capabilities through text analysis of annual reports and apply fixed-effects regression models to test our hypotheses. The results show that higher AI capability significantly improves ESG performance. Mechanism analysis suggests that AI enhances ESG outcomes by optimizing resource allocation and increasing efficiency in production and supply chains. Further, the positive effect of AI on ESG performance is more pronounced in industries with intense competition, while it is weakened under high environmental uncertainty. These findings contribute to the growing literature on AI and corporate sustainability by revealing both the internal mechanisms and contextual contingencies that shape ESG performance. The study offers practical insights for corporate managers aiming to leverage AI for sustainable development and provides policy recommendations for fostering supportive external environments.

1. Introduction

The rapid progress of artificial intelligence (AI) has significantly transformed the global economic landscape. In particular, AI technologies—such as cloud computing, machine learning, intelligent systems, deep learning, cognitive networks, and natural language processing—have increasingly been integrated into business operations to enhance strategic decision-making and improve organizational performance [1]. As a critical force of the Fourth Industrial Revolution, AI, with its powerful computational capabilities, precise data analysis, and intelligent learning algorithms, endows enterprises with unprecedented insights and adaptability. This technology not only automates burdensome processes but also plays a transformative role in forecasting market trends, optimizing resource allocation, and facilitating decision-making, thereby significantly enhancing a firm’s core competitiveness and value creation capacity [2].

According to PwC’s 2023 report (The report is sourced from the AI Adoption in the Business World: Current Trends and Future Predictions 2023, PwC. Available at: https://www.pwc.com/il/en/mc/ai_adopion_study.pdf, accessed on 29 December 2024), AI applications currently used in businesses are primarily focused on decision support, efficiency improvement, revenue growth, cost control, and customer service optimization. Additionally, data indicates that approximately 30% of enterprises in China are exploring AI applications, compared to over 45% in the United States. These figures underscore the significant trends and growing importance of AI in the corporate domain, highlighting its gradual emergence as a pivotal tool for driving innovation and sustainable development.

However, while AI technology helps firms pursue efficiency and competitive advantages, it also presents challenges in balancing economic objectives with social responsibilities amidst the growing global consensus towards sustainable development. Firms must not only stand out in market competition but also demonstrate strong performance in ESG. Therefore, investigating how firms leverage AI to enhance competitive advantages while improving ESG performance is key to fostering long-term corporate growth and advancing societal sustainability.

With increasing interest in practice, academic research has also achieved significant progress in exploring how AI technology supports corporate sustainability. Scholars have approached this topic from various perspectives. Bahoo et al. pointed out that AI enables companies to manage big data and information, thereby fostering corporate innovation [3]. Chen et al. emphasized that AI technology can help firms mitigate financing constraints and lower agency costs, thereby enhancing environmental governance and social responsibility [4]. Ganesh and Kalpana, from a risk management perspective, highlighted that AI applications help companies alleviate supply chain risks, thus advancing corporate sustainability [5]. While these studies offer valuable insights into the diverse benefits of AI in corporate operations and governance, they predominantly emphasize its practical applications and thus leave open the important question of whether and how AI adoption influences corporate ESG performance. This gap limits a more comprehensive understanding of AI’s strategic role in promoting long-term sustainability. AI possesses the capacity to enhance corporate operational efficiency, raise investment in research and development, and optimize resource allocation. This allows redundant resources to be redirected towards other corporate practices. As ESG represents a long-term corporate investment, the efficiency gains from AI may enable companies to dedicate more resources to fulfilling their ESG commitments. Therefore, exploring the influence and mechanisms of AI on ESG performance is essential for advancing corporate sustainability.

To address the above issues, this paper uses A-share listed companies on the Shanghai and Shenzhen stock exchanges in China from 2010 to 2023 as the initial sample. It constructs an analytical model to explore the relationship between AI and corporate ESG performance. The study examines how AI enhances ESG performance through improved operational efficiency and optimized resource allocation. Furthermore, this paper examines the potential impact of external environmental factors, highlighting their significant effect on corporate performance.

Our study makes three main contributions: First, it extends the research on how AI influences corporate performance and the underlying mechanisms. While prior studies have examined AI’s influence across various domains, such as corporate governance [6], corporate failure prediction [7], innovation efficiency [8], and consumer value creation [9], limited studies have been conducted on whether and how AI influences corporate ESG performance. Therefore, this paper, combining the resource-based view, stakeholder theory, and transaction cost theory, systematically explores the impact of AI on ESG performance. Furthermore, it delves into the process mechanisms underlying this effect, revealing the key role of corporate efficiency as an important mediator. Specifically, we investigate how AI enhances operational efficiency and supply chain effectiveness, optimizes resource allocation, and ultimately improves ESG performance. This finding enriches the theoretical framework on how AI improves corporate sustainability performance and offers a new research perspective on the social responsibility implications of AI usage.

Second, this paper broadens the study of factors determining corporate ESG performance. Prior research has investigated the impact of managerial characteristics [10], industry traits [11], and corporate globalization [12] on ESG performance, but few studies focus on how AI technology applications at the corporate level drive improvements in ESG performance. As AI use rapidly increases, its impact on reshaping corporate competitive dynamics is increasingly significant. AI enhances corporate innovation capabilities and operational efficiency, alters the competitive landscape, and optimizes resource allocation and decision-making quality, thereby improving corporate performance in ESG aspects. Thus, analyzing ESG performance solely from corporate characteristics or industry perspectives is no longer sufficient or adaptable to the rapidly evolving business environment. By focusing on AI technology applications, this study systematically explores their far-reaching impact on corporate ESG performance, broadening the research path to improving ESG outcomes and providing a new theoretical framework and practical guidance for corporate social responsibility research.

Finally, this paper focuses on corporate behavior patterns in industries with high competition and environmental uncertainty, emphasizing the importance of external contexts in explaining performance differences. Pagell and Krause argue that external factors significantly affect a firm’s operational flexibility [13], while Navarro-García et al. emphasize that strategic positioning is fundamental to a company’s success in complex and dynamic external environments [14]. When external conditions are volatile, such changes can significantly impact key financial indicators, such as revenue and investment returns. Thus, this paper adopts an external context perspective to explore the effects of AI on ESG performance, offering new theoretical insights into corporate adaptation and performance in dynamic and complex environments. This analysis not only deepens research on the influence of external environments on business performance but also provides important academic and practical guidance for formulating strategies that ensure sustainable growth in an evolving environment.

2. Theoretical Framework and Hypothesis Development

2.1. AI and ESG Performance

The resource-based view (RBV) theory argues that a firm gains a competitive advantage through its unique resources and capabilities, which are essential, scarce, difficult to replicate, and cannot be easily replaced [15]. As an emerging and cutting-edge resource, AI demonstrates strong technological potential and is also essential in ESG practices. First, from the environmental perspective, AI can facilitate breakthroughs in fields such as energy management [16], emissions control [17], and resource optimization [18]. In response to increasingly stringent environmental regulations and societal expectations, firms must dynamically monitor the environmental impact of their production activities. Leveraging its robust data analytics capabilities, AI enables real-time collection, processing, and analysis of relevant data, providing precise environmental management solutions for firms. These solutions not only help reduce emissions and pollution-related costs but also enhance corporate reputation in environmental responsibility [19]. Second, in the context of social responsibility, AI enhances firms’ responsiveness to stakeholder needs, thereby promoting the fulfillment of corporate social responsibilities. Firms face diverse demands from consumers, communities, and employees, and the application of AI allows them to identify and address these needs more efficiently [20]. For instance, firms can employ AI to monitor social responsibility issues within supply chains, such as labor conditions or supplier behaviors, and take timely corrective actions. This improves their ability to meet social responsibility obligations while strengthening trust with external stakeholders, resulting in stronger performance in the social responsibility dimension. Finally, the role of AI in corporate governance cannot be overlooked. Firms face complex managerial challenges in governance, such as risk control and compliance audits. Through its efficient algorithms and intelligent tools, AI technology enhances transparency and accuracy in governance processes [21]. For instance, AI assists firms in real-time monitoring of operational risks and predicting potential issues, thereby strengthening governance capabilities and compliance performance. This technological empowerment directly improves governance efficiency and provides crucial support for the governance dimension of firms’ overall ESG performance.

According to stakeholder theory, the integration of AI fosters deeper collaboration and trust between companies and their stakeholders, thereby enhancing corporate ESG performance. Stakeholder theory suggests that businesses must coordinate and address the diverse expectations of stakeholders, which is critical for strengthening corporate legitimacy and sustainability [22]. Within this framework, AI technology provides an effective tool for optimizing and managing these relationships, driving improvements in ESG outcomes. Specifically, AI enhances transparency and traceability, helping firms align their actions more closely with stakeholder expectations in the ESG domain [23]. With AI, companies can efficiently collect data, conduct real-time analysis, and generate reports that more accurately demonstrate their progress in ESG practices. This mechanism of transparency and real-time feedback not only boosts corporate credibility but also enables firms to respond promptly and effectively to stakeholder needs and expectations, particularly in the areas of environmental and social responsibility. Furthermore, AI demonstrates unique strengths in addressing environmental changes [4]. When facing pressures from societal, environmental, or regulatory sources, AI can assist companies in dynamic adjustments and resource optimization, ensuring that they fulfill social responsibilities while maintaining stable stakeholder relationships. These advantages position AI not just as a tool for enhancing ESG performance but as a means of strengthening stakeholder interactions and collaboration, thereby providing robust support for corporate sustainability.

Building on the previous analysis, our study presents the following hypothesis:

H1. 

The use of AI is positively correlated with corporate ESG performance.

2.2. The Mediating Mechanism of Production Efficiency and Supply Chain Efficiency

According to transaction cost theory, companies face various internal and external transaction costs, such as information search costs, negotiation costs, and the costs of contract execution and monitoring, in their pursuit of economic efficiency [24]. To minimize these costs, firms often seek to optimize resource allocation and improve operational processes, thereby enhancing competitiveness and achieving more efficient production and supply chain systems. AI technology is crucial to this process as it significantly boosts production and supply chain efficiency [25]. First, AI enhances production efficiency by automating production processes [26]. In traditional production models, firms typically rely on manual operations and decision-making, which may result in prolonged production cycles, higher error rates, and resource wastage [27]. The implementation of AI technology automates numerous aspects of production, not only improving efficiency but also enhancing product quality. For instance, AI-driven robots can perform complex manufacturing tasks, and computer vision technologies can conduct quality inspections [28]. More importantly, AI enables instant collection and assessment of production metrics, allowing firms to identify bottlenecks and inefficiencies, thereby facilitating more refined production management. These capabilities effectively increase production efficiency, freeing up resources for investment in other areas. Second, AI technology has a profound impact on optimizing supply chain efficiency. The supply chain is a core component of enterprise operations, with its efficiency directly influencing cost control and market responsiveness [29]. Traditional supply chain management often encounters challenges such as information asymmetry, demand forecasting errors, and improper inventory management, all of which can lead to resource wastage and capital constraints [30]. Through data analysis and optimization algorithms, AI can accurately predict market demand, enabling firms to adjust production and inventory plans, thereby avoiding excessive stockpiling and improving liquidity [31]. Additionally, AI can streamline logistics routes and distribution networks, reducing transportation costs and enhancing the entire supply chain efficiency. These technological advancements not only improve supply chain responsiveness but also strengthen firms’ ability to adapt to market fluctuations and unforeseen events [32].

By enhancing production efficiency and supply chain efficiency, AI strengthens firms’ resource allocation capabilities, enabling them to operate more effectively in increasingly complex and uncertain external environments. The reduction in operating costs and improvement in resource allocation efficiency allow firms to free up resources, which can then be reinvested into long-term initiatives, particularly ESG investments. According to transaction cost theory, effective resource allocation helps firms mitigate uncertainty and friction in the external environment, thereby improving their market flexibility and competitiveness [33]. In the ESG domain, one of the greatest challenges for firms lies in the delayed returns on investment, particularly in areas such as environmental technology development, the implementation of social responsibility initiatives, and the optimization of corporate governance structures, which often require sustained funding and resource commitment with limited short-term financial returns [34]. However, by improving production and supply chain efficiency, AI enables firms to reallocate the resources saved into these long-term investment areas, providing stable financial support for ongoing improvements in ESG performance. AI not only helps firms reduce short-term operating costs but also optimizes internal workflows and external supply chain management through automation and intelligent tools, further improving resource utilization efficiency. This enhanced efficiency extends beyond immediate economic benefits, creating greater capacity for fulfilling social and environmental responsibilities. Consequently, firms can channel more resources into initiatives that may not yield instant financial benefits but are critical for their long-term development. By reducing reliance on short-term profits, firms can consistently support their ESG objectives over an extended period, ultimately improving their long-term social responsibility performance.

Building on the previous analysis, the following hypotheses are presented:

H2. 

AI enhances a firm’s ESG performance through improved production efficiency.

H3. 

AI enhances a firm’s ESG performance through improved supply chain efficiency.

2.3. Moderating Mechanisms of Industry Competitiveness and Environmental Uncertainty

2.3.1. Moderating Mechanisms of Industry Competitiveness

The effectiveness of AI in enterprises is not reflected in a single dimension, but is influenced by a variety of external factors. Industry competitiveness, as a crucial characteristic of the external market environment, significantly impacts corporate strategic decisions and performance [35]. In highly competitive industries, companies must seek more effective strategic drivers under resource constraints to maintain competitiveness in an intense market environment [36]. AI technology, with its significant advantages in data processing, resource optimization, and decision support, provides an essential tool for improving ESG performance. As competition intensifies, companies’ focus on ESG shifts from passive compliance to proactive optimization, with AI technology becoming an indispensable tool in this context. In this process, the level of industry competitiveness significantly enhances the depth and effectiveness of AI in ESG practices, particularly in industries with strong sustainability and social responsibility demands. AI enables companies to gain more accurate insights and execution capabilities in ESG management.

Moreover, in industries with high competitiveness, the effects of benchmarking and imitation behavior on corporate actions becomes more pronounced [37]. When a company significantly enhances its ESG performance through AI and gains market recognition, competing firms often imitate this approach, further driving the widespread integration of AI across the industry. This dynamic competitive mechanism not only broadens the application of AI but also elevates the overall ESG standards within the industry. As competition intensifies, companies increasingly recognize that improving ESG performance not only provides compliance advantages but also strengthens brand reputation and attracts investors, thereby advancing the industry’s overall sustainability goals [38].

Finally, as competition intensifies, the external oversight on corporate behavior, particularly regarding ESG performance, also increases significantly [39]. Stakeholders demand higher levels of transparency and accountability from companies, and AI can assist businesses in meeting these demands by providing precise data analysis and efficient information management. The stronger external pressure in a competitive environment further motivates companies to adopt AI, making it a crucial tool for enhancing ESG performance. Through technological empowerment, businesses not only improve their reputation and influence in the market but also establish a more sustainable competitive advantage in a highly competitive market [40]. AI enhances short-term competitiveness while simultaneously establishing a solid foundation for long-term ESG performance, enabling businesses to sustain a market-leading position.

Building upon the analysis presented above, this study formulates the following hypothesis:

H4. 

Industry competitiveness strengthens the promoting effect of AI on ESG performance.

2.3.2. Moderating Mechanisms of Environmental Uncertainty

Environmental uncertainty, as another crucial feature of the external environment, equally influences corporate strategic choices and performance outcomes [41]. On one hand, under high environmental uncertainty, companies face an external environment filled with instability factors such as policy changes, market fluctuations, and other unpredictable elements, which causes them to prioritize short-term survival and risk mitigation [42]. In such contexts, businesses often adjust their strategies by prioritizing resources to address the short-term risks and operational pressures caused by these uncertainties. This shift in focus leads to a reduction in attention to long-term strategic goals, especially regarding the implementation of AI in improving ESG performance. ESG optimization typically requires continuous investment of resources and time, but in high uncertainty environments, companies become skeptical about the returns from long-term investments. Consequently, they tend to concentrate resources on areas that can quickly stabilize market position and enhance current competitiveness [43]. Although AI technologies can help businesses make improvements in sustainable development, the strategic shift towards short-term, return-driven actions diminishes AI’s role in enhancing ESG performance, especially when resources are limited and market conditions are unstable.

On the other hand, in environments with high uncertainty, companies not only face external fluctuations but also need to contend with internal strategic uncertainties [44]. The implementation of AI often requires complex technological integration, data accumulation, and system optimization, all of which are typically long-term and resource-intensive processes. In such an environment, companies may be more inclined to focus on current operational efficiency and short-term returns, rather than recognizing the potential of AI to promote ESG practices. High environmental uncertainty leads companies to question the complexity and costs associated with implementing AI, particularly when clear returns cannot be guaranteed in the short term. As a result, companies may delay or reduce investments in AI for ESG goals [45]. The implementation of AI typically requires cross-departmental collaboration and long-term strategic planning, which conflicts with the strategic choices companies tend to make in high-uncertainty environments [46]. Consequently, the contribution of AI in improving ESG performance is diminished, with companies relying more on decisions that ensure short-term survival and stability. Meanwhile, the risk-averse tendencies of businesses may lead to more cautious investments in AI, further slowing the pace of AI in promoting ESG performance.

Building on this, the following hypothesis is presented:

H5. 

Environmental uncertainty weakens the positive impact of AI on ESG performance.

The conceptual framework of this research is shown in Figure 1.

3. Research Method

3.1. Sample Selection and Data Sources

This study utilizes all A-share listed companies on the Shanghai and Shenzhen Stock Exchanges in China from 2010 to 2023 as the primary sample. Following established research practices and design requirements [47], the initial sample was refined through the following steps: excluding companies with negative net assets; removing companies under special treatment (ST), particularly *ST and PT companies (i.e., firms with abnormal financial conditions or delisting risks as designated by the stock exchanges); excluding firms in the financial sector; and eliminating observations with missing values for the research variables. To minimize the impact of outliers, all continuous variables were winsorized at the 1st and 99th percentiles. The measurement of corporate AI level is based on text analysis of annual reports. ESG scores were retrieved from the Shanghai Huazheng database. Unless otherwise specified, all other data were derived from the China Stock Market & Accounting Research (CSMAR) database.

3.2. Model Design

This study develops the following regression model to analyze the effect of AI on ESG performance, incorporating year-fixed effects and industry-fixed effects. Additionally, the model incorporates various firm-level variables (Controls) to enhance the reliability of the empirical analysis. In addition, standard errors are adjusted for heteroskedasticity to improve the reliability of statistical inference under potential model misspecification.

$$ES{G}_{i,t}={\alpha }_0+{\alpha }_{1}A{I}_{i,t}+\Sigma {\alpha }_{j}Control{s}_{i,t}+\Sigma Year+\Sigma Indus+{\epsilon }_{i,t}$$

(1)

Building upon the analytical framework proposed by Edwards and Lambert [48], which systematically integrates mediation and moderation analyses within a regression-based approach, we utilize the regression models outlined below to examine Hypotheses 2 and 3, exploring the mediating roles of corporate production and supply chain efficiency. TFP and SCE represent the production efficiency and supply chain efficiency of firm i in year t, respectively.

$$TF{P}_{i,t}={\beta }_0+{\beta }_{1}A{I}_{i,t}+\Sigma {\beta }_{j}Control{s}_{i,t}+\Sigma Year+\Sigma Indus+{\epsilon }_{i,t}$$

(2)

$$ES{G}_{i,t}={\beta }_0+{\beta }_{1}A{I}_{i,t}+{\beta }_{2}TF{P}_{i,t}+\Sigma {\beta }_{j}Control{s}_{i,t}+\Sigma Year+\Sigma Indus+{\epsilon }_{i,t}$$

(3)

$$SC{E}_{i,t}={\beta }_0+{\beta }_{1}A{I}_{i,t}+\Sigma {\beta }_{j}Control{s}_{i,t}+\Sigma Year+\Sigma Indus+{\epsilon }_{i,t}$$

(4)

$$ES{G}_{i,t}={\beta }_0+{\beta }_{1}A{I}_{i,t}+{\beta }_{2}SC{E}_{i,t}+\Sigma {\beta }_{j}Control{s}_{i,t}+\Sigma Year+\Sigma Indus+{\epsilon }_{i,t}$$

(5)

Subsequently, we use the following models to analyze the moderating effects of industry competitiveness and environmental uncertainty. IC represents the level of industry competitiveness for firm i in year t, and EU denotes the level of environmental uncertainty.

$$ES{G}_{i,t}={\chi }_0+{\chi }_{1}A{I}_{i,t}+{\chi }_{2}A{I}_{i,t}\ast I{C}_{i,t}+\Sigma {\chi }_{j}Control{s}_{i,t}+\Sigma Year+\Sigma Indus+{\epsilon }_{i,t}$$

(6)

$$ES{G}_{i,t}={\chi }_0+{\chi }_{1}A{I}_{i,t}+{\chi }_{2}A{I}_{i,t}\ast E{U}_{i,t}+\Sigma {\chi }_{j}Control{s}_{i,t}+\Sigma Year+\Sigma Indus+{\epsilon }_{i,t}$$

(7)

3.3. Measures

3.3.1. Independent Variable

Corporate AI level (AI): Based on the approach adopted by Liu et al. [49], this study utilizes machine learning techniques to generate an artificial intelligence dictionary and constructs a measure of corporate AI level based on the annual reports of samples. Specifically, 68 keywords, including “cloud computing”, “machine learning”, “intelligent system”, “artificial intelligence”, “deep learning”, “cognitive networks”, “natural language processing”, “business intelligence”, “blockchain”, and “automated systems”, were selected as seed words. Using Python 3.9, the study performs text segmentation and analysis of corporate annual reports. The frequency of AI-related keywords in the reports is then calculated, and the natural logarithm of the frequency, after adding one, is used to construct the measure of corporate AI level.

3.3.2. Dependent Variable

Corporate ESG performance (ESG): We use ESG scores provided by the Shanghai Huazheng database to measure corporate ESG performance. There are two primary reasons for this choice. First, Huazheng’s ESG rating system offers advantages in terms of its long timespan and broad coverage compared to existing rating systems, ensuring an adequate sample size for regression analysis. Second, Huazheng’s ESG rating is renowned for its scientific rigor and has been widely recognized and applied by both industry practitioners and academic researchers, ensuring high reliability and validity.

3.3.3. Mediating Variables

Production efficiency (PE): The study adopts total factor productivity as a key variable to assess corporate production efficiency. TFP not only reflects technological progress but also encompasses factors such as the knowledge level in material production, managerial capabilities, institutional environment, and potential measurement errors, providing a complete assessment of AI’s role in enhancing production efficiency [50]. Therefore, this study estimates TFP using the LP method introduced by Levinsohn and Petrin [51], as shown in the following formula. In the calculation process, the total output variable in TFP is measured as the natural logarithm of a firm’s revenue; labor input is represented by the natural logarithm of workforce size; capital input is represented by the natural logarithm of net fixed assets; and intermediate input is represented by the natural logarithm of the total of operating costs, management expenses, selling expenses, and financial costs, excluding depreciation, amortization, and cash payments to employees.

$$\ln Y_{i,t}={\alpha }_0+{\alpha }_1\ln L_{i,t}+{\alpha }_2\ln K_{i,t}+{\alpha }_3\ln M_{i,t}+{\omega }_{i,t}$$

(8)

Supply chain efficiency (SCE): Supply chain efficiency is reflected not only in the overall improvement of profits but also in the operational cycle from a time dimension perspective [52]. This research emphasizes the time aspect of supply chain efficiency, drawing upon the research of Hwang et al. [53], and employs the net operating cycle as a primary measure. The net operating cycle is derived by combining the accounts payable period, inventory turnover period, and accounts receivable period. This indicator effectively reflects the operational efficiency of the procurement, production, and sales stages within the supply chain. The net operating cycle of core firms is closely related to the synergy between upstream and downstream companies, with shorter cycles indicating higher supply chain efficiency.

3.3.4. Moderating Variables

Industry competitiveness (IC): This study measures industry competitiveness using the inverse of the Herfindahl–Hirschman Index (HHI). This indicator reflects market concentration and competition intensity by calculating the inverse of total squared market shares across all companies in the sector. A higher IC value indicates lower market concentration and higher competition, while a lower IC value suggests higher market concentration and lower competition. Specifically, the market share of each company is represented by the proportion of its main business revenue to the overall industry revenue. The HHI is computed using the following formula, where s_i represents the market share of firm i, and N represents the number of firms in the industry.

$$HHI={\displaystyle \sum _{i=1}^N{s}_i^2}$$

(9)

Environmental uncertainty (EU): Following the approach proposed by Ghosh and Olsen [54], this study quantifies environmental uncertainty by calculating the industry-adjusted standard deviation of a firm’s sales revenue over the past five years. Specifically, the volatility of a firm’s sales revenue reflects the extent of external environmental fluctuations and uncertainty it faces. Greater volatility indicates that the firm is encountering more significant challenges in a turbulent and unstable market environment. To adjust for industry effects, the study first calculates the sales revenue fluctuations of the firm over the past five years, then compares it with the standard deviation of sales revenue fluctuations of other firms in the same industry. This comparison controls for general market fluctuations within the industry, yielding the firm-specific environmental uncertainty indicator.

3.3.5. Controls Variables

Based on the work of Kulkov et al. [55], this study controls for firm-specific factors including firm size (Size), debt ratio (Lev), return on equity (ROE), years since listing (ListAge), the shareholding ratio of the top ten shareholders (Top10), duality (Dual), and the number of board members (Board). The detailed description of these variables is presented in

Table 1. Variable definition.

Variable	Variable Definition
AI	The frequency of artificial intelligence keywords in the annual reports of sample companies, adjusted by adding 1 and then applying the natural logarithm
ESG	ESG scores from Huazheng database
PE	The natural logarithms of output, labor, capital inputs, and intermediate inputs
SCE	Measured by the net operating cycle, which is calculated by combining the accounts payable period, inventory turnover period, and accounts receivable period
IC	Measured using the inverse of Herfindahl–Hirschman Index
EU	The industry-adjusted standard deviation of a firm’s sales revenue over the last five years
Size	The natural logarithm of total assets
Lev	Total debts/total assets
ROE	Measured as net income divided by shareholders’ equity
Growth	Difference between current period and previous period revenue/pervious period revenue
ListAge	The length of time from its initial public offering (IPO) to the current year
Top10	The proportion of shares held by the top 10 shareholders to the total shares outstanding
Dual	A binary variable that equals 1 if the chairman and CEO are the same person, and 0 otherwise.
Board	The logarithmic value of the number of board members in the company

.

4. Empirical Results

4.1. Descriptive Statistics

Table 2. Descriptive statistics.

Variables	N	Mean	SD	Min	Median	Max
ESG	46,865	73.1384	5.008	57.39	73.35	84.23
AI	46,865	0.8541	1.196	0.00	0.00	4.54
PE	46,865	8.3196	1.065	6.03	8.23	11.20
SCE	46,865	4.4860	1.289	−0.28	4.56	7.79
IC	46,865	0.0821	0.074	0.01	0.06	0.37
EU	46,865	1.3612	1.221	0.14	1.00	7.30
Size	46,865	22.2253	1.443	19.60	21.98	27.27
Lev	46,865	0.4314	0.220	0.05	0.42	0.95
ROE	46,865	0.0588	0.144	−0.74	0.07	0.37
Growth	46,865	0.1585	0.422	−0.59	0.10	2.70
ListAge	46,865	2.0524	0.932	0.00	2.20	3.37
Top10	46,865	0.5842	0.156	0.23	0.59	0.91
Dual	46,865	0.2840	0.451	0.00	0.00	1.00
Board	46,865	2.1234	0.206	1.61	2.20	2.71

displays the descriptive statistics for the relevant variables. As shown in Table 2, the average ESG performance is 73.1384, with a standard deviation of 5.008, reflecting significant variability across the companies. The companies’ average AI level is 0.8541, but the median is 0, implying that more than half of the companies have AI levels below the average. The mean, standard deviation, minimum, and maximum values for firm production efficiency are 8.3196, 1.065, 6.03, and 11.20, respectively, showing that there is little variation in production efficiency among the sample companies, with values remaining within a certain range. The mean of supply chain efficiency is 4.4860, with a maximum of 7.79, a minimum of −0.28, and a standard deviation of 1.289, suggesting that there is considerable variation in supply chain efficiency across the firms, possibly resulting from differences in resource allocation and management practices. The mean industry competitiveness level for the sample companies is 0.0821, with a standard deviation of 0.074, indicating relatively small overall variation in industry competitiveness among the sample firms. Regarding environmental uncertainty, the average value for the sample companies is 1.3612, with a maximum of 7.30 and a minimum of 0.14, suggesting significant differences in environmental uncertainty among the firms, which may reflect the varying external environmental challenges faced by different companies.

4.2. Correlation Analysis

Table 3. Correlation analysis.

	ESG	AI	PE	SCE	IC	EU	Size
ESG	1
AI	0.138 ***	1
PE	0.239 ***	0.070 ***	1
SCE	0.032 ***	0.049 ***	0.194 ***	1
IC	−0.037 ***	−0.018 ***	0.042 ***	−0.106 ***	1
EU	−0.201 ***	−0.044 ***	−0.100 ***	0.044 ***	0.004	1
Size	0.266 ***	0.032 ***	0.291 ***	−0.069 ***	0.028 ***	−0.071 ***	1
Lev	−0.106 ***	−0.083 ***	0.393 ***	−0.017 ***	0.038 ***	0.069 ***	0.383 ***
ROE	0.231 ***	−0.038 ***	0.219 ***	−0.047 ***	−0.016 ***	−0.105 ***	0.109 ***
Growth	−0.017 ***	−0.018 ***	0.116 ***	−0.058 ***	0.014 ***	0.371 ***	0.028 ***
ListAge	−0.139 ***	−0.093 ***	0.277 ***	−0.004	0.025 ***	0.111 ***	0.345 ***
Top10	0.175 ***	−0.039 ***	0.148 ***	−0.075 ***	0.023 ***	−0.039 ***	0.132 ***
Dual	0.005	0.120 ***	−0.119 ***	0.050 ***	−0.045 ***	−0.016 ***	−0.182 ***
Board	0.066 ***	−0.097 ***	0.173 ***	−0.056 ***	0.050 ***	−0.057 ***	0.315 ***
	Lev	ROE	Growth	ListAge	Top10	Dual	Board
Lev	1
ROE	−0.221 ***	1
Growth	0.032 ***	0.245 ***	1
ListAge	0.382 ***	−0.162 ***	−0.045 ***	1
Top10	−0.134 ***	0.234 ***	0.082 ***	−0.229 ***	1
Dual	−0.165 ***	0.018 ***	0.012 ***	−0.249 ***	0.040 ***	1
Board	0.191 ***	0.055 ***	0.003	0.120 ***	0.028 ***	−0.184 ***	1

Note: *** report the significance level at 1%.

presents the correlation coefficients among the variables. Table 3 illustrates that the correlation between AI and ESG performance is 0.138, significant at the 1% level. This provides preliminary support for the core hypothesis of our study, specifically that AI positively influences ESG performance. Moreover, the absolute values of the correlation coefficients among the variables do not exceed 0.5, implying that multicollinearity is not a considerable issue in this study.

4.3. Baseline Regression Results Analysis

To identify the impact of AI on ESG performance, this study employs a panel data regression framework with industry and year fixed effects. This method is well-suited to control for unobserved time-invariant heterogeneity across industries and common shocks over time, thereby improving the reliability of causal inference. The estimation relies on standard assumptions underlying linear panel data models, which are appropriately addressed through the model specification and estimation approach.

This study applies Equation (1) to examine the relationship between AI and ESG performance, with the detailed results presented in

Table 4. Baseline regression results and robustness test results.

Variable	(1)	(2)	(3)	(4)
	ESG	ESG	ESG	ESG
AI	0.5465 ***	0.3729 ***
	(23.17)	(17.29)
AI_1			0.4119 ***
			(17.42)
AI_2				0.4570 ***
				(19.53)
Size		1.4352 ***	1.4035 ***	1.4344 ***
		(69.05)	(62.88)	(68.68)
Lev		−4.7784 ***	−4.6963 ***	−4.7387 ***
		(−37.85)	(−34.72)	(−37.42)
ROE		4.3752 ***	7.0801 ***	4.3911 ***
		(27.23)	(41.19)	(27.23)
Growth		−0.6742 ***	−0.0047	−0.6667 ***
		(−13.28)	(−0.09)	(−13.08)
ListAge		−0.9778 ***	−0.7413 ***	−0.9791 ***
		(−33.31)	(−23.60)	(−33.25)
Top10		−0.0276	−0.2003	−0.0210
		(−0.17)	(−1.18)	(−0.13)
Dual		−0.1049 **	−0.1388 ***	−0.1025 **
		(−2.21)	(−2.71)	(−2.15)
Board		−0.1953 *	−0.0712	−0.2010 *
		(−1.81)	(−0.62)	(−1.85)
Constant	72.6917 ***	45.3443 ***	44.9853 ***	45.3465 ***
	(241.91)	(106.66)	(98.93)	(106.06)
Year/Industry	Yes	Yes	Yes	Yes
N	46,865	46,865	42,239	46,865
Adj R2	0.0661	0.2272	0.2483	0.2254

Note: ***, **, * report the significance level at 1%, 5%, and 10%, respectively.

. Column (1) displays the empirical results with only year and industry fixed effects included, while Column (2) includes the relevant control variables. The coefficients for AI are 0.5465 and 0.3729, with both showing statistical significance at the 1% level. These results demonstrate that AI enhances corporate ESG performance, thereby validating Hypothesis 1.

4.4. Robustness Tests

First, this study considers the possibility of a lagged effect of AI on ESG performance, which may not be fully reflected in the ESG outcomes of the current year. Consequently, all independent (AI_1) and control variables are lagged by one period for robustness testing. Additionally, the measurement of AI is refined. Given that the management discussion and analysis (MD&A) section of corporate annual reports often provides critical insights into business operations, financial performance, R&D activities, and strategic planning, this study incorporates the MD&A section as an alternative proxy. Specifically, the natural logarithm of the frequency of AI-related keywords within the MD&A section, incremented by one (AI_2), is employed as a new measure of corporate AI level to re-examine Hypothesis 1.

The findings presented in columns (3) and (4) of Table 4 indicate that the coefficients for AI are 0.4119 and 0.4570, with both being statistically significant at the 1% level. These findings from the robustness checks further confirm Hypothesis 1, demonstrating that AI promotes ESG performance.

4.5. Endogeneity Analysis

4.5.1. Instrumental Variable (IV) Regression

This study adopts the instrumental variable approach to mitigate potential endogeneity concerns, utilizing the number of AI companies in the province where the sample firms are based as the instrument for corporate AI level. The rationale for choosing this variable is based on its substantial impact on the AI level of local firms [56], while it is unlikely to have a direct influence on the ESG performance of individual companies. This fulfills the conditions of relevance and exogeneity for an instrumental variable.

Table 5. Endogeneity test results.

	(1)	(2)	(3)	(4)
	AI	ESG	ESG_dummy	ESG
IV	0.0431 ***
	(20.01)
AI		1.7826 ***		0.1836 ***
		(12.12)		(5.87)
IMR				1.3253 ***
				(2.67)
Size	0.0974 ***	1.2833 ***	0.0217	1.3418 ***
	(20.88)	(49.35)	(0.92)	(31.14)
Lev	−0.3901 ***	−3.8317 ***	−0.1166	−4.9352 ***
	(−13.18)	(−26.33)	(−0.82)	(−28.00)
ROE	−0.2875 ***	5.0471 ***	−0.1672 ***	0.0051
	(−7.09)	(25.05)	(−2.69)	(0.32)
Growth	0.0558 ***	−0.7536 ***	0.0808	−0.0007 **
	(4.65)	(−12.56)	(1.34)	(−2.46)
ListAge	−0.1908 ***	−0.6512 ***	1.2367 ***	−0.9357 ***
	(−27.86)	(−15.48)	(29.93)	(−8.77)
Top10	−0.9861 ***	1.3044 ***	−0.2070	−0.7556 ***
	(−25.29)	(5.89)	(−0.98)	(−2.91)
Dual	0.1412 ***	−0.3025 ***	0.0118	−0.0095
	(11.00)	(−5.44)	(0.28)	(−0.15)
Board	−0.1678 ***	−0.0280	−0.1856	−0.5047 ***
	(−6.14)	(−0.23)	(−1.64)	(−2.97)
Constant	−0.5663 ***	47.5128 ***	1.5102 ***	48.7222 ***
	(−6.22)	(108.84)	(3.16)	(50.89)
Year/Industry	Yes	Yes	Yes	Yes
N	46,253	46,253	45,905	45,905
Adj R2	0.1670	0.1108	0.4156	0.4711

Note: ***, ** report the significance level at 1% and 5%, respectively.

summarizes the results of the two-stage least squares (2SLS) regression analysis. In the first stage, displayed in column (1), the instrumental variable demonstrates a coefficient of 0.0431, which is statistically significant at the 1% level. In the second stage, shown in column (2), the coefficient for AI is 1.7826, also achieving significance at the 1% level. These results further validate Hypothesis 1, confirming that corporate AI level positively influences ESG performance.

4.5.2. Heckman Two-Stage Model

In China, the disclosure of ESG reports is largely voluntary, except for certain industries. As a result, the Huazheng ESG rating database may suffer from incomplete data, which could introduce sample selection bias into this study’s findings. To address this issue, the Heckman two-stage model is utilized. First, a binary variable is created for ESG performance, where a value of 0 is assigned if ESG performance data is missing, and 1 otherwise. A first-stage regression is then conducted based on this binary variable, using the same control variables as outlined previously. Second, the inverse Mills ratio (IMR) is derived from the probit model regression and added as a control variable to the baseline regression model. The results are illustrated in columns (3) and (4) of Table 5.

The results indicate that the regression coefficient for AI is 0.1836, statistically significant at the 1% level. Similarly, the IMR coefficient is 1.3253 and significant at the 1% level, confirming the presence of sample selection bias in this analysis. Even after addressing sample selection bias, the positive association between AI and ESG performance remains statistically significant, further validating Hypothesis 1. These results strengthen the robustness of the study’s conclusions.

4.6. Analysis of the Mediating Mechanism

To explore the mediating effects of corporate production and supply chain efficiency, this study utilizes equations (2) through (5), with the results displayed in

Table 6. Mediating mechanism analysis.

	(1)	(2)	(3)	(4)
	PE	ESG	SCE	ESG
AI	0.0488 ***	0.3685 ***	0.0019 ***	0.3699 ***
	(16.43)	(16.00)	(6.63)	(16.91)
PE		0.3181 ***
		(8.37)
SCE				0.0748 ***
				(3.91)
Size	0.5723	1.2982 ***	−0.0098 *	1.4413 ***
	(99.25)	(41.79)	(−1.89)	(68.17)
Lev	0.7476 ***	−5.112 ***	−0.2972 ***	−4.7850 ***
	(43.19)	(−37.41)	(−9.48)	(−37.55)
ROE	1.0931 ***	3.8913 ***	−0.3856 ***	4.3195 ***
	(50.43)	(22.57)	(9.68)	(26.69)
Growth	0.1251 ***	−0.7065 ***	−0.1618 ***	−0.6444 ***
	(17.76)	(−12.93)	(−12.71)	(−12.44)
ListAge	0.0223 ***	−1.0918 ***	−0.0366 ***	−0.9878 ***
	(4.73)	(−29.95)	(−4.90)	(−32.54)
Top10	0.2242 ***	−0.2288	−0.4707 ***	0.0328
	(10.26)	(−1.35)	(−11.80)	(0.20)
Dual	−0.0214 ***	−0.1567 ***	0.0717 ***	−0.1082**
	(−3.25)	(−3.08)	(6.10)	(−2.26)
Board	−0.0633 ***	−0.2424 **	0.0581 **	−0.3234 ***
	(−4.22)	(−2.09)	(2.15)	(−2.94)
Constant	−4.8613	46.3862 ***	5.0690 ***	45.1278 ***
	(−82.37)	(94.33)	(47.43)	(101.52)
Year/Industry	Yes	Yes	Yes	Yes
N	46,865	46,865	46,865	46,865
Adj R2	0.7137	0.2163	0.3098	0.2139

Note: ***, **, * report the significance level at 1%, 5%, and 10%, respectively.

. Columns (1) and (2) test the mediating role of corporate production efficiency. The AI coefficient in column (1) is 0.0488, which is statistically significant at the 1% level, indicating that AI significantly enhances corporate production efficiency. Column (2) further examines the impact of production efficiency on ESG performance, showing a coefficient of 0.3181, also significant at the 1% level. This finding suggests that improved efficiency significantly enhances ESG performance. Collectively, the empirical results confirm the mechanism by which AI indirectly promotes ESG performance through increased corporate production efficiency, supporting Hypothesis 2.

The mediating effect of supply chain efficiency is explored in columns (3) and (4). The results in column (3) indicate that the AI coefficient is 0.0019 and significant at the 1% level, suggesting that AI significantly improves supply chain efficiency. The results in column (4) further demonstrate a favorable relationship between supply chain efficiency and ESG performance, showing a coefficient of 0.0748, statistically significant at the 1% level. These results validate the mechanism by which AI enhances ESG performance through improved supply chain efficiency, thus supporting Hypothesis 3.

4.7. Analysis of the Moderating Mechanism

We further examine the moderating effects of industry competitiveness and environmental uncertainty, as displayed in

Table 7. Moderating mechanism analysis.

	(1)	(2)
	ESG	ESG
AI	0.2398 ***	0.4515 ***
	(8.15)	(15.44)
AI×IC	1.7875 ***
	(6.34)
IC	−2.4977 ***
	(−5.69)
AI×EU		−0.0725 ***
		(−4.82)
EU		−0.4450 ***
		(−20.08)
Size	1.4389 ***	1.3909 ***
	(69.23)	(66.46)
Lev	−4.7766 ***	−4.8301 ***
	(−37.85)	(−38.23)
ROE	4.3712 ***	3.5973 ***
	(27.22)	(22.03)
Growth	−0.6774 ***	−0.0447
	(−13.35)	(−0.79)
ListAge	−0.9741 ***	−0.9193 ***
	(−33.18)	(−30.55)
Top10	−0.0060	0.1458
	(−0.04)	(0.92)
Dual	−0.1084 **	−0.1083 **
	(−2.28)	(−2.28)
Board	−0.1951 *	−0.3147 ***
	(−1.81)	(−2.91)
Constant	45.4402 ***	46.9414 ***
	(106.87)	(108.80)
Year/Industry	Yes	Yes
N	46,865	46,865
Adj R2	0.2280	0.2376

Note: ***, **, * report the significance level at 1%, 5%, and 10%, respectively.

. The coefficient for the interaction term between AI and industry competitiveness in column (1) is 1.7875, which is statistically significant at the 1% level. This suggests that in industries with higher competition, AI has a more pronounced positive impact on ESG performance, thereby supporting Hypothesis 4. The second column displays the results for environmental uncertainty, with the interaction term coefficient being −0.0725, also significant at the 1% level. This indicates that environmental uncertainty weakens the positive effect of AI on ESG performance, thus confirming Hypothesis 5.

5. Discussion

This study provides valuable insights into the relationship between AI and corporate ESG performance. Based on empirical results, it uncovers the mediating mechanisms of production efficiency and supply chain efficiency, as well as the moderating roles of industry competition and environmental uncertainty. In this section, we discuss the theoretical and managerial implications of these findings.

5.1. Theoretical Implications

This study offers several key theoretical contributions to the intersection of artificial intelligence and corporate sustainability research.

First, this study extends the understanding of how AI contributes to corporate ESG performance. While previous studies have established that AI enhances innovation capability [3], organizational flexibility [57], goal achievement [58], and R&D investment [59], few have examined its role in driving ESG performance. This paper fills that gap by empirically demonstrating that AI significantly improves ESG outcomes. It also enriches the theoretical conversation by introducing production efficiency and supply chain efficiency as key explanatory mechanisms. These findings provide a novel theoretical framework that links digital transformation with corporate sustainable development, thereby advancing scholarship on the integration of emerging technologies and ESG practices.

Second, this study contributes to the broader literature on the determinants of ESG performance by highlighting the role of digital technologies. While prior studies have emphasized traditional management factors such as board diversity [60], shareholder activism [61], and executive incentives [62], they have largely overlooked the strategic influence of AI. This study suggests that AI, as a long-term investment in digital capability, significantly reshapes how firms allocate resources, manage operations, and pursue innovation, all of which are critical for achieving sustainability goals. Thus, the findings expand the current theoretical landscape by positioning AI as a strategic enabler of ESG performance.

Third, this study introduces external environmental factors to explore the differences in corporate performance under varying external conditions, highlighting the substantial influence of the external environment on corporate performance. Sukumar et al. suggest that an increase in industry competitiveness helps enhance a company’s innovation capability to maintain its competitive advantage [63]. This finding indicates that evolving external environment has a profound influence on corporate decision-making. Our study further demonstrates that industry competitiveness enhances AI’s positive impact on ESG, thus helping companies maintain competitiveness and market share. On the other hand, environmental uncertainty, as another key external factor, affects not only financial performance [64] and strategic decisions [41] but also plays a moderating role in the relationship between AI and ESG performance. In cases of high environmental uncertainty, companies tend to reduce long-term investments to address short-term survival pressures. This finding enhances the understanding of how external environmental factors affect business performance and provides both academic insights and practical recommendations for companies to develop sustainable strategies in dynamic and complex environments.

5.2. Managerial Implications

This study also offers several managerial insights that can guide firms in integrating AI to enhance ESG performance.

First, companies should place significant emphasis on the development of AI capabilities. As a cutting-edge technology aligned with modern technological trends and innovation needs, AI holds considerable strategic value and long-term influence [65]. With the accelerating global digital transformation, AI technology can not only deliver economic benefits in the short term by enhancing production efficiency and innovation but also promote sustainable performance in the ESG domain in the long term. On one hand, companies should strengthen investment in AI research and development and integrate it into their overall strategic planning. Through continuous technological innovation and in-depth application, companies can effectively improve operational efficiency, optimize resource allocation, and enhance their long-term competitiveness in ESG while ensuring short-term economic benefits. On the other hand, companies should focus on cultivating management and technical teams with high levels of AI expertise. Establishing cross-departmental collaboration mechanisms and promoting the deep integration of AI technology in aspects such as production, supply chain management, and strategic decision-making will provide strong technical support for the company’s sustainable development efforts.

Second, companies should prioritize enhancing production and supply chain efficiency, as this is essential for maximizing the benefits of AI and boosting ESG performance. Specifically, companies can establish and strengthen information-sharing platforms to enable real-time monitoring and information visibility across the supply chain, ensuring that all parties can respond quickly and adjust strategies in response to market changes. In addition, companies should build long-term strategic partnerships with upstream and downstream supply chain partners to share technological innovations and data resources, jointly driving improvements in production and supply chain efficiency. Through these measures, companies can not only enhance operational efficiency and reduce cost risks but also better apply AI in production and supply chain system, laying a robust base for the long-term development of ESG objectives.

Third, while focusing on their own development, companies should also prioritize on the shifts in the external environment, particularly the impact of industry competitiveness and environmental uncertainty on their strategic decisions. In the context of increasing industry competitiveness and growing environmental uncertainty, companies must not only address short-term market competition and external shocks but also develop proactive long-term strategies. Specifically, companies should establish a robust competitive intelligence system to monitor industry dynamics, market trends, and changes in competitors in real time, helping them identify potential competitive threats and opportunities. At the same time, companies should conduct risk assessments and adjust resource allocation strategically to maintain a competitive advantage in the intense market competition. Furthermore, companies should enhance their operational flexibility and adaptability to better cope with external uncertainties. Ultimately, companies must develop strategies to navigate the evolving external environment and secure long-term sustainable success in response to intense competition and market uncertainty.

6. Conclusions and Limitations

6.1. Conclusions

This study empirically investigates the impact of AI on ESG performance using a comprehensive dataset of Chinese listed firms from 2010 to 2023. The results indicate that AI significantly enhances ESG performance by improving production and supply chain efficiency. Furthermore, external environmental factors play a crucial moderating role. While industry competitiveness amplifies AI’s positive effect on ESG, environmental uncertainty weakens it, as firms prioritize short-term survival over long-term investment. These findings enrich theoretical discourse at the intersection of AI and corporate sustainability and provide meaningful insights for firms seeking to integrate emerging technologies with sustainable development goals.

6.2. Limitations and Future Research

This study analyzes the influence of AI on ESG performance and provides both theoretical and empirical support. However, the study also presents some limitations and suggests avenues for future research. First, this study measures AI using a text analysis approach applied to corporate annual reports. While this method is widely used to assess AI capabilities [25,49], it should be acknowledged that AI levels involve various aspects of a company, including its technological foundation, R&D investment, and the utilization of AI technologies in different scenarios. Therefore, relying solely on data from annual reports may not fully demonstrate the AI level of a company. Future research could attempt to employ more refined and multidimensional analytical methods, such as incorporating data on R&D investments, technological innovations, and AI application cases, to comprehensively assess a company’s AI level from various dimensions. Second, this study mainly uses cross-industry data for empirical analysis, but the effect of AI on ESG performance may differ significantly across industries. For instance, AI adoption in high-tech industries may primarily enhance innovation-driven ESG outcomes, whereas in traditional manufacturing, the focus may be more on resource efficiency and emission reduction. Future research could explore these industry-specific mechanisms by conducting comparative studies between sectors, thereby offering a more contextualized understanding of how AI influences ESG performance in different institutional and operational environments. Third, although our study examines the direct relationship between AI and ESG performance, the long-term effects of AI on ESG have not been sufficiently validated. Given that AI adoption is often a long-term strategic investment, its influence on sustainability outcomes may not be immediately observable. Moreover, AI may shape a firm’s ESG strategy differently at various stages of its development, such as during digital transformation, organizational restructuring, or sustainability transitions. Future research could adopt longitudinal designs to examine the dynamic impact of AI on ESG over time, uncovering how these effects evolve and persist across different phases of corporate growth.