Unveiling the mechanisms and implications: how artificial intelligence drives green growth in China’s Huaihe River Ecological Economic Belt under the carbon neutrality agenda

Green innovation effect

Green innovation develops slowly due to the problem of high initial investment and lengthy payback duration. Still, the development of AI provides a breakthrough, accelerating technological breakthroughs by improving an innovation-friendly environment and technology externalities and gradually becoming the core driving force to promote environmental governance and green development^[39]. First is the enabling role of AI technology. Chen et al. find that AI can empower green innovation, and this enabling effect shows nonlinear characteristics of decreasing empowerment range after a marginal increase to a certain extent^[31]. The opposition between economic expansion and environmental preservation is alleviated by optimizing resource allocation and use and strengthening the environment’s fine management. In the meantime, new solutions are provided to mitigate problems involving dynamic environmental pollution surveillance, and the government’s intelligent management ability is enhanced. Bera et al. found through empirical research that cutting-edge technologies such as smart grids and energy-saving technologies can also be used to reduce production energy consumption and refine energy utilization^[40]. Open-source green innovation initiatives frequently lack viable commercial frameworks, struggling to secure sustained R&D investment. This financial instability results in suboptimal algorithm refinement for critical systems like smart grid management and energy-saving technologies, ultimately hindering advancements in resource-use efficiency and decarbonization efforts. Second is the role of AI technology innovation. AI promotes the efficient flow of green innovation resources, breaks through the limitation of spatial boundaries, expands the breadth of resource allocation, helps innovation subjects integrate resources across time and space, and expands green innovation’s potential and development space. Tang et al. also proved this conclusion^[41]. Proprietary AI, dominated by corporate data and algorithm monopolies, restricts green innovation resource sharing. Analyzing AI’s green development mechanisms via green innovation elucidates the merits and drawbacks of both proprietary and open AI. Moreover, through the masking effect of technology spillover, it strengthens the role of improving green total factor productivity (TFP), stimulates innovation efficiency, optimizes factor allocation, promotes the improvement of green TFP, and finally achieves the aspiration for a sustainable and resilient future.

Structural adjustment effect

Within the sphere of AI, the strategic optimization of the economic structure not only holds the potential to unlock structural dividends but also serves as a powerful catalyst for the swift and transformative shift of the economy toward a green and carbon-efficient trajectory. By leveraging the advanced capabilities of AI, we can meticulously analyze complex economic patterns, identify inefficiencies, and implement targeted interventions that drive sustainable growth. This intelligent reconfiguration of resources promises to revolutionize traditional industries, fostering a harmonious integration of technological innovation and ecological stewardship. Ultimately, it paves the way for a symbiotic future where economic prosperity and environmental sustainability coexist, heralding a new era of responsible development. First, AI can optimize resource allocation. AI can substantially uplift traditional industries’ production efficiency and market competitiveness by enhancing the intelligence level of conventional industries. Wang et al. demonstrated through empirical evidence that AI can promote the sophisticated calibration of the occupational structure of medium-skilled labor and highly skilled labor, and there is a nonlinear relationship with the employment structure of labor^[42]. However, the impact of AI on the labor market is twofold. Bonfiglioli et al. found that, to a certain extent, by replacing part of the low-end labor force, the labor force is promoted to develop in the direction of more interpersonal interaction, complex task processing, and flexible adaptation to optimize resource allocation^[23]. If workers who have been replaced cannot transition, they may fall into technological unemployment. Therefore, we should consider this aspect when exploring the mechanisms. Second, AI can promote industrial upgrading. With the traits of deep penetration, data-oriented, and intelligent architecture integration, Li et al. have demonstrated that AI not only encourages the technological novelty of traditional commercial sectors but also promotes the growth of human capital, thus improving the industrial structure and fundamentally enhancing the optimization of China’s industrial structure^[25]. In summary, Figure 1 illustrates how the development of AI impacts the green development of the HEB through technology-enabling and technology spillover, resource allocation optimization, and industrial structure upgrading.

Figure 1. The effect of AI technology on the green transformation of the ecological economic belt. AIT: Artificial intelligence technology.

Modeling

A baseline regression model was first established to assess AI’s ramifications on HEB’s green development. Before establishing the model, Stata 16.0 software was used to judge and analyze the panel data preliminarily. The panel data in this paper were tested as balanced panel data, and the Hausman test results showed that the fixed effect model was adopted. To provide additional substantiation for the model’s applicability, the joint significance test of all year dummy variables is undertaken, and the data underscore that the model has both time and regional effects. Given the inherent individual heterogeneity and temporal trends in panel data, the two-way fixed effects model controls individual and time-fixed effects simultaneously. This approach effectively mitigates the influence of time-invariant urban characteristics and individual-invariant temporal shocks, enabling precise identification of the causal relationship between AI and green development. The Hausman test rejects the random effects model. In contrast, the joint significance test confirms significant temporal effects, thus justifying the adoption of the two-way over the one-way fixed effects model. Drawing from the preceding analysis, the bidirectional fixed-effect model is selected to control time and prefecture-level city variables to effectively eliminate the potential repercussions of heteroscedasticity on the results derived from the model. Despite controlling for various confounders, unobservable omitted variables may introduce bias. Future research could use instrumental variables or natural experiments to address endogeneity. The dual fixed effect model is set as follows:

In formula (1), GEDI_it represents the green development stratum of the city i within the temporal framework of year t; AI_it represents the AI composite development profile of city i in the temporal context of year t; ∑Control_it represents a repertoire of control variables, including per capita GDP, industrial structure, scientific and technological level, opening-up degree, and urbanization rate.

Variable selection

Explained variables

Urban green development level of HEB (GEDI). From the actual development status of 27 cities in the HEB, since the typical resource allocation model permits the principle of efficiency maximization, this decentralized and flexible weight will cause some decision-making units (DMUs) to ignore specific inputs or outputs to attain maximum efficiency, which will lead to unfair resource distribution. Moreover, the traditional data envelopment analysis (DEA) model may lead to inconsistent conclusions due to different weight choices when evaluating different DMUs. Therefore, the traditional DEA model cannot fully show the differences and comparability of green development in cities, while the common set of weights DEA (CSW-DEA) model can solve this problem by a unified weight set.

Therefore, this study refers to the general weight calculation method proposed by Feng et al., namely the CSW-DEA model, to measure the index of the GEDI^[43]. CSW-DEA model is a new DEA model based on the principle of non-egoism. This model has a weight set (CSW) for fixed cost allocation. Common weights are derived from the collective input and output of all DMUs and applied across them. Twenty-seven prefecture-level resource-based cities are selected as the research sample, and each prefecture-level city serves as a DMU with consistent input and output types. The K DMU consumes m different inputs x_ik (i = 1,2,…,m) and produces n different outputs y_jk (j = 1,2,…,n). The input indicators include the stock of fixed capital, total energy consumption, and the number of employees in each region, and the output indicators include the expected output GDP, the pollutants of atmospheric carbon dioxide concentrations, wastewater effluents from industrial processes, and sulfurous by-products of industrial activities, which are not expected output. The efficiency level of any given DMUk is expressed in formula (2):

(2) $$ \begin{array}{c} \max e_{k}=\frac {\displaystyle \sum_{j=1}^{n} u_{j}^{k} y_{j k}}{\displaystyle\sum_{i=1}^{m} v_{i}^{k} x_{i k}}\\ { s.t. }\left\{\begin{array}{l} \frac {\displaystyle\sum_{j=1}^{n} u_{j}^{k} y_{j k}} {\displaystyle\sum_{i=1}^{m} v_{i}^{k} x_{i k}} \le 1, \forall k \\ u_{j}^{k}, v_{i}^{k} \geq 0, \forall i, j, k \end{array}\right. \end{array} $$

Where e_k is the efficiency score of DMU_k, $$ u_{j}^{k} $$ and $$ v_{i}^{k} $$ are the weights of the jth output and the ith input of DMU_k, correspondingly. The expression of formula (2) is nonlinear, and it is more convenient to solve the overall efficiency and its corresponding common weight by transforming it into a linear form, such as formula (3).

(3) $$ \begin{array}{l} \max e_{k}= \displaystyle \sum_{j=1}^{n} u_{j}^{k} y_{j k} \\ { s.t. }\left\{\begin{array}{l} \displaystyle \sum_{i=1}^{m} v_{i}^{k} x_{i k}=1 \\ \displaystyle \sum_{j=1}^{n} u_{j}^{k} y_{j k}-\displaystyle \sum_{i=1}^{m} v_{i}^{k} x_{i k} \leq 0, \forall k \\ u_{j}^{k}, v_{i}^{k} \geq 0, \forall i, j, k \end{array}\right. \end{array} $$

Assuming that ($$ u_{j}^{*}, v_{i}^{*} $$) is the best collective efficiency, that is, the optimal solution, the input-output volume of DMU_k is exhibited in formula (4):

(4) $$ S_{k}=\sum_{j=1}^{n} u_{j}^{*} y_{j k}+\sum_{i=1}^{m} v_{i}^{*} x_{i k}, \forall k $$

Thus, the green development level of the 27 cities in the HEB can be precisely calculated. As visualized in Figure 2, the data from 2022 are taken as an example to analyze the divergence in the green development levels of all provinces in the HEB. Notably, the region is confronted with the dual challenges of a relatively backward level of high-quality development and pronounced regional imbalances in green development. This underscores the necessity for a deeper and more intricate comprehension of the foundational dynamics and the development of targeted strategies to address these disparities.

Figure 2. Digital frequency chart of GEDI in 2022. GEDI: Green development level of Huaihe Ecological Economic Belt.

Explanatory variables

AIDL: Breaking from traditional variable selections in prior studies, this document adopts the logarithm of the number of AI firms as a metric representing the AI development level, following the approach of Babina et al.^[44]. Through the Python crawler technology, the fuzzy matching of keywords such as “intelligence,” “cloud,” “data,” and “Machine Intelligence Learning” is carried out to gather the multi-dimensional data of AI enterprises in each urban region from 2010 to 2022. The logarithm of the extent of AI enterprises in the city that year is used as a diagnostic tool to judge the local AI development level. This paper selects the AIDL in 2010 (the 9th Five-Year Plan), 2015 (the 11th Five-Year Plan), 2018 (proposed by the HEB), and 2022 for visualization processing. As shown in Figure 3, a chart depicting AI development levels over the four years was created to enable a comparative evaluation of the measured levels across these regions for the duration of the study. In light of the overall trend, the spatial distribution shows unbalanced characteristics. Throughout the entire period, Jiangsu Province consistently exhibits a higher level of AI than the other three provinces. Meanwhile, Anhui Province has the lowest development level. It is speculated that this may be closely related to the economic growth level and geographical location.

Figure 3. AIDL of the HEB in 2010, 2015, 2018, and 2022. AIDL: Artificial intelligence development level; HEB: Huaihe River Ecological Economic Belt.

Baseline regression results

This study adopted the bidirectional fixed-effect model for regression, and the coefficients between variables were below 0.7. The variance inflation factor (VIF) values corresponding to each variable were below 5 to exclude the possibility of multicollinearity among variables. The results showed that explanatory and control variables did not show significant multicollinearity.

Table 2 shows the baseline regression results with control variables added step by step. When the model is specified with only temporal and regional fixed effects and control variables are excluded, the coefficient of the core explanatory variable on AIDL emerges as significantly positive, achieving statistical significance at the 1% level. This underscores the robust influence of AIDL in the initial model configuration, highlighting its pivotal role in driving the observed outcomes. After gradually introducing control variables, the key coefficient remains positive at the 1% significance level. And passed the significance test, indicating that the AIDL can promote the green development of HEB.

Integrating control variables significantly enhances the constructive effect of AI on the GEDI. The negative coefficients of industrial structure, urbanization rate, and openness to the outside world indicate that capital preferentially flows to AI industries. In contrast, excessive urbanization and openness may exacerbate environmental pollution. The positive coefficient of GDP per capita supports the Kuznets curve hypothesis that environmental conditions will improve after a particular stage of economic resurgence. The progress of scientific and technological endeavors has no discernible effect on green development. AI improves resource efficiency, reduces pollution, promotes coordinated regional economic growth, and provides technical support for sustainable development in various ways.

Robustness test

The robustness test is conducted on the baseline regression results to ensure the correctness of the model estimation outcomes. Specific methods are as follows:

(1) Changing the regression model: This paper selects the Tobit model for a robustness test to oversee the sample selection bias caused by the dependent variable being restricted by certain constraints. It can be seen that the use of Tobit regression does not change the soundness of the empirical estimates, and the AIDL will still substantially boost the GEDI.

(2) Winsorization: To mitigate the potential distortion of regression results caused by outliers, the dependent variables in this study are subjected to a 1% Winsorization procedure. The corresponding results are presented in Column (2).

(3) Interactive fixed effects: Considering that there are unobservable factors that fluctuate across various time points at the provincial level, the joint fixed effects of time × province and region × province are further controlled based on the foundational regression model employed as a baseline, and the end products are depicted in Columns (3) and (4).

(4) Lagged explanatory variable: Considering the lagged effect of the development of AI on the GEDI, the AIDL, the principal explanatory variable, is subjected to a one-period temporal displacement.

(5) Sample Time Adjustment: Considering that the core explanatory variables may have measurement bias during the epidemic, the sample time is adjusted, and only the data from 2010 to 2020 are retained for analysis, and the results are shown in Column (6).

(6) Sub-sample regression: four cities of Yancheng, Huai’an, Bengbu, and Xinyang in the sustainable development initiatives nexus of the downstream region of the downstream territories of the Huaihe River are selected to observe the most straightforward effect of AIDL on the GEDI, and the results are shown in Column (7).

Table 3 shows the results of the robustness test. As shown in the table, the robustness test results collectively reaffirm the conclusion that AI functions as a pivotal and foundational element in underpinning the environmental sustainability of the HEB.

Test for endogeneity

To ameliorate the potential endogeneity issues, this study draws upon the Two-stage least squares Estimation Method for testing. The mean value of AIDL in other cities in the same province (IV1) and the log value of AIDL enterprise stock (IV2) are selected as instrumental variables. From columns (1) and (3) in the table, it is evident that the two instrumental variables have successfully passed both the weak instrument test and the underidentification test, and both are markedly positive at the level of 1%. Table 4 shows the regression results of the endogeneity test. The regression coefficients in columns (2) and (4) are still significantly positive at 1%. In panel data, traditional instrumental variables (IV) may not effectively deal with the bias caused by the correlation between individual fixed effects and lagged terms. Therefore, after using instrumental variable regression to address the endogeneity issue, the system generalized method of moments (GMM) approach is further employed for endogeneity testing to enhance the robustness and credibility of the estimation. The regression results of the system GMM model show that the first-order lag term of the explanatory variable x and the explained variable y both have a significant positive impact on y, and the model as a whole is significant. The results above reveal that, within the sample period under examination, the salutary influence of AI on the GEDI remains significantly pronounced, even after the potential endogenous biases have been meticulously mitigated through the application of the instrumental variable method. This underscores the robust and enduring positive correlation between AI advancements and the trajectory of green economic development, reinforcing that AI catalyzes sustainable progress in this domain.

Test of the mechanism

Acting as a crucial catalyst for a fresh epoch of scientific and technological disruption and industrial metamorphosis, AI has significantly improved the effectiveness of urban green development and brought about a beneficial transformation in the green development of the ecological, economic enclave through its green technology innovation effect and industrial structure adjustment effect. To investigate the influence of AI on GEDI further, this study uses the mediating effect model to test whether AI affects GEDI by acting on green innovation and structural adjustment. Table 5 shows the baseline regression results of the mechanism test.

First, in the quest to explore through empirical testing the green innovation effect of AI, this paper calculates the number of green and low-carbon technology patent applications in each region derived from the “GLTPCS” prepared by the State Intellectual Property Office. It uses the occurrence of green patent applications per 10,000 individuals (GPA) to represent the quality of regional green inventiveness. Drawing on the approach of Du et al., the density of green patent grants per 10,000 population (GPG) was used to characterize the number of regional green innovations to illustrate the complexity of green technology innovation^[45]. Columns (1) to (4) represent the green innovation effect, and columns (5) to (8) represent the structural adjustment effect. The outcomes listed in columns (1) and (3) show that AIDL has a decisive and transformative impact on green technology innovation. A significant negative interconnection exists when testing the association between core explanatory variables and intermediary variables. In contrast, when testing explained variables, intermediary variables, and core explanatory variables, there is a significant positive correlation, indicating a U-shaped correlation between AI and green technology innovation. In its nascent stages of development, AI demands substantial amounts of energy, thereby increasing carbon emissions. However, energy utilization efficiency improves as technological advancements unfold, reducing carbon emissions. The experiment conducted by Chen et al. shows that both the direct and spillover effects reveal that only when green technology innovation reaches a certain threshold can it positively influence environmental improvement and carbon emission control^[46]. Governments and relevant authorities should take proactive measures to ensure that each city’s green innovation level is positioned on the right side of the “inverted U-shaped” curve, thereby fully leveraging the positive impact of green technology. The results presented in columns (2) and (4) highlight that the coefficient levels for GPA and GPG are 0.002 and 0.003, respectively, implying that the evolution of AI has substantially driven the improvement and increment of green innovation output and promoted the green and eco-friendly and sustainable development of the economic belt through the effect of green innovation.

Secondly, to quantify the structural adjustment effects of AI, two critical variables have been meticulously selected: Ais and Thile. Empirical evidence from columns (5) and (7) reveals that AIDL exerts a profound and transformative impact on structural adjustment. Conversely, the results from columns (6) and (8) indicate that Ais, while significant, fail to reject the null hypothesis at the 10% significance level. In stark contrast, Thile emerges as significantly positive at the 1% significance level, thereby underscoring AI’s burgeoning and progressive role in catalyzing structural upgrading and rationalization adjustment. This suggests that, as a high-tech industry, AI has the potential to recalibrate and optimize the architectural integrity of secondary and tertiary industry frameworks. Ultimately, this recalibration is poised to have a more favorable impact on GDP, paving the way for sustainable and innovative economic growth. Figure 4 shows that AI drives the green development of the Huai River Ecological Economic Belt via green innovation quality, quantity, and industrial structure upgrading & rationalization.

Figure 4. Action path of AI. AI: Artificial intelligence.

Research conclusions

In economic transformation, AI, as a crucial determinant that holds sway over the outcome to promote the formation of a new zenith of optimized efficiency and productivity, has great potential to guide China’s economic expansion force and achieve the vision of enduring growth and environmental stewardship. This paper starts from a new perspective of new quality productivity, and uses the cross-sectional time-series data of 27 cities in the HEB from 2010 to 2022. The empirical test explored the catalytic role and operational framework of v on the green development of HEB. The overarching conclusions are elucidated as follows: First, the advancement in AI has profoundly catalyzed the green development trajectory of the Huaihe region. This assertion maintains its validity even when accounting for potential endogenous biases. Second, AI exhibits a uniform impact across diverse temporal and spatial contexts, thereby underscoring its consistent influence. Third, through the conduits of green innovation and structural adjustment, AI exerts a salutary influence on the green development of the Huaihe region. This demonstrates that AI is a crucial catalyst in propelling green technological advancements and facilitating the transition of industries toward an environmentally sustainable structure. Optimizing the structure of industrial sectors and enhancing inter-sectoral coordination can boost economic efficiency and sustainability, facilitating the transition of industries from traditional high-pollution models to AI-integrated green industries. Although this transformation may cause short-term job losses in traditional sectors, it will generate new employment opportunities in AI-driven fields and green industries. This study elucidates the ramifications of AI on the green development of the Huaihe region, deepens the insight into the complex interaction between AI and green development, and reveals the latent role of AI in promoting sustainable economic transformation. This is achieved in tandem with the impetus derived from environmental protection and sustainable development.

Countermeasures and suggestions

First, vigorously develop the green effect of AI. It is essential to unleash AI’s green potential vigorously across the HEB. Enhancing the maturity of AI applications within diverse regions of the HEB and continuously tapping its potential to drive low-carbon, efficient, and climate-resilient development are critical steps. Integrating AI technology with environmental monitoring and protected area management is a key driver for enhancing environmental governance and promoting green development. Yu et al. used a deep ensemble learning model with global data from 5,446 stations to predict PM2.5 concentrations at 0.1° × 0.1° resolution, reaching 91% accuracy^[47]. Chen et al. achieved 83% accuracy in China and the US with a Bayesian-random forest PM2.5 model^[48]. Lu et al. improved river water quality prediction precision by combining CEEMDAN and XGBoost^[49]. Integrating AI technology into environmental monitoring and reserve management can significantly elevate the precision and efficiency of environmental governance, thereby continuously stimulating the impact of AI on green development. Leveraging DeepMind’s AI algorithms to forecast renewable energy variability, California, USA, integrated this technology with its renewable energy quota system. This approach reduced wind and solar power curtailment rates from 15% to 5%, substantially improving power grid stability. Moreover, the deep integration of AI with the green industry must be promoted to drive the transformation of the industrial structure toward a green and low-carbon trajectory. This will help mitigate carbon emissions and reduce industrial activities with adverse environmental repercussions. Additionally, the impact mechanisms through which AI influences the green development of the HEB require further exploration. It is crucial to fully leverage AI’s guiding and driving role to accelerate the innovation and development of intelligent industrial technologies and green-related technologies.

Second, the government should restructure the intelligent manufacturing ecosystem and innovation framework to fortify the industrial structure. The government needs to persistently foster the Scientific and Technological development of v and other eco-friendly technologies to catalyze industrial modernization further and ensure that AI maximizes its effectiveness in advancing green development. Drawing on the success of ThyssenKrupp Steel in Germany, the company utilized AI to optimize blast furnace gas ratios in real-time. Coupled with the EU’s “green steel” tax relief policies, it achieved a 22% reduction in carbon emissions per ton of steel, resulting in an annual emission reduction of 500,000 tons. Initially, it is imperative to catalyze the intelligent transformation of traditional industries expeditiously. By leveraging the advanced capabilities of AI, we can judiciously curtail energy consumption and emissions, propelling the manufacturing sector toward a paradigm characterized by high-end innovation, intelligent operations, and environmental benignity. This transformation is a technological upgrade and a strategic pivot toward sustainable industrial practices. Subsequently, there must be a significant augmentation of financial expenditure allocated to the research and development initiatives of AI-driven green technologies. This includes robust support for related scientific research projects and the encouragement of enterprises to engage actively in green technology innovation. By doing so, we can accelerate the diffusion of green technologies and enhance their market penetration. This involves uplifting existing facilities’ energy-saving and carbon-reduction transformation, guiding the green and low-carbon development of digital technology enterprises, and assisting both upstream and downstream enterprises in enhancing their carbon-reduction capabilities. In conclusion, the paramount imperative lies in prioritizing green and low-carbon digital infrastructure construction. This endeavor is not merely a strategic choice but a fundamental necessity in the contemporary era, where the confluence of technological advancement and environmental sustainability defines the trajectory of global progress. By emphasizing the development of such infrastructure, we can ensure that principles of ecological responsibility and long-term viability underpin our digital future. This approach will foster innovation and economic growth and contribute to the broader goal of mitigating environmental impact, thereby aligning our technological ambitions with the imperatives of a sustainable planet. AI’s green potential is universal across industrial, energy, agricultural, and ecological management domains. Its success relies on the synergy of technology, policy, and localization. Developed regions leverage high-quality data and market mechanisms for rapid technological diffusion, while developing countries must strengthen infrastructure and skill training. Compared with the HEB, international cases highlight the efficacy of government-led models in cross-sectoral resource coordination, yet also emphasize the necessity of flexible policies to enhance social acceptance. A global open-source platform for AI green technologies could foster cross-regional collaboration, standardize solutions, and enable regionally tailored innovation. Through these concerted efforts, we can create a resilient and sustainable digital ecosystem that aligns with global environmental goals.

Thirdly, leveraging the mediating effect as a catalyst for indirect advancement is imperative. With its multi-faceted and intricate impacts, AI drives structural adjustments across multiple dimensions, including resource allocation, industrial integration, emerging industry cultivation, production efficiency enhancement, and informed decision-making. This process reshapes the economic landscape of the Huaihe River Ecological Economic Belt. By fostering green technology innovation and optimizing industrial structures, AI plays a pivotal, albeit indirect, role in advancing green development. While automation may diminish the demand for labor in routine tasks, it concurrently spurs the need for high-skilled positions. This shift underscores the imperative of implementing labor retraining programs to address skill mismatches. Such dynamic interactions highlight the pressing need for the government to catalyze the transformation of environmental protection industries and technological breakthroughs. The balanced distribution of green industries within the HEB promotes regional equity and mitigates development disparities. Such a vision can be realized by deploying robust financial sponsorship, strategic tax incentives, and establishing collaborative platforms that foster synergies among stakeholders. Moreover, governments should champion the international diffusion and joint sharing of green technologies, encouraging a more sustainable and interconnected world. Governments should promote the global dissemination and sharing of green technologies through international cooperation, innovation platforms, green financial support, and the Belt and Road Initiative to advance global sustainable development. Additionally, they should enhance international corporate collaboration, education, and training to expand the application and adoption of green technologies worldwide.

Future research endeavors may delve deeper into several pivotal directions. Expanding the scope of research regions and conducting comparative analyses of the green development effects of AI across diverse economic and environmental contexts could significantly enhance the generalizability of research conclusions. Secondly, an extensive scrutiny of the specific application pathways of AI technologies within various industries and their differential impacts on green transitions would provide more targeted policy recommendations. Thirdly, integrating big data and machine learning methodologies to monitor the long-term impacts of AI on green development dynamically could unveil the intricate interplay between AI and socio-economic factors. Future research may also focus on the role of AI in catalyzing global green transitions, fostering international cooperation and technology sharing, and offering theoretical and practical guidance for the realization of global sustainable development goals.