Impact of artificial intelligence on the achievement of sustainable development goals across China

Introduction: The rapid development of artificial intelligence (AI) technology has become a key driver in achieving the Sustainable Development Goals (SDGs). However, systematic quantitative researches about the impact and pathway of artificial intelligence on the achievement of SDGs are still relatively limited.

Methods: Based on the penal data of 30 provinces (autonomous regions and municipalities) in China from 2010 to 2023, this study uses the entropy method to calculate evaluation index of SDG indicators, and systematically examines the impact and pathway of artificial intelligence (measured by the installation density of industrial robots) on achieving SDGs through the two-way fixed effect model, the mediation effect model, and spatial error model.

Results: The study finds that artificial intelligence has markedly promoted the achievement of China’s overall SDGs and it has a continuously increasing cumulative effect. This driving effect varies across different SDGs, and is most significant in economic growth (SDG8 and SDG9) and climate action (SDG13). The effect of artificial intelligence on achieving SDGs is more pronounced in eastern regions, regions with better SDGs performance, and regions with a late start in artificial intelligence development. Technological innovation and human capital optimization are important ways through which artificial intelligence affects the achievement of SDGs and can further enhance its positive role. Spatial analysis reveals significant positive spatial agglomeration in both AI and SDGs, along with potential spatial dependence across regions driven by unobserved factors.

Discussion: This study provides useful decision-making references for formulating differentiated and inclusive artificial intelligence policies, and promoting the achievement of SDGs.

1 Introduction

In recent years, artificial intelligence (AI) technology has witnessed explosive development worldwide, with continuous breakthroughs in fields such as natural language processing, computer vision, and machine learning. From early symbolic logic to today’s large deep learning models, artificial intelligence has achieved a leap from basic scientific research to industrial transformation, permeating almost every aspect of human life and driving unprecedented progress in various fields, including communication, healthcare, engineering, and sustainable development (Jordan and Mitchell, 2015; Merz et al., 2024). As a key driver of the Fourth Industrial Revolution (Dwivedi et al., 2021), artificial intelligence has attracted the attention of the academic community, become a focus for technology and service enterprises, and is continuously promoting social progress and economic transformation (Goralski and Tan, 2020).

As global climate change challenges intensify and natural resources continue to deplete, sustainable development has attracted widespread attention from the international community (Jiang and Ma, 2025). In 2015, the United Nations established 17 Sustainable Development Goals (SDGs) and 169 specific targets at its Development Summit, aiming to call on multiple stakeholders, including countries, organizations and individuals, to take action to jointly address challenges in the social, economic, and environmental fields, and provide an action framework and directional guidance for the global transition to sustainable development (Colglazier, 2015). China incorporated the SDGs into its national development strategy in 2016, yet still faces numerous challenges in the process of implementation. Against this backdrop, how to fully leverage the advantages of artificial intelligence development, explore effective pathway for AI technology to empower sustainable development, accelerate the implementation of the SDGs in China and globally, and thereby achieve higher-level goal attainment has become a critical challenge that urgently needs to be addressed.

Since the establishment of the SDGs, academic interest in sustainable development has increased significantly. Early studies primarily focused on the social, economic, and environmental dimensions of sustainable development, and explored the factors affecting the realization of SDGs and the construction of SDGs indices (Diaz-Sarachaga et al., 2018). With the development of artificial intelligence technology, people are beginning to realize its great potential in promoting SDGs, especially in providing data-driven insight and optimization processes (Gohr et al., 2025). In recent years, a growing body of literature has explored the connection between artificial intelligence and SDGs, indicating that artificial intelligence may have both positive and negative consequences for sustainable development. In the economic dimension, artificial intelligence can promote economic growth (SDG 8) (Zhao et al., 2022), drive industrial upgrading (SDG 9) (Rammer et al., 2022; Xia et al., 2024), and expand international trade (DeStefano and Timmis, 2024). In the social dimension, artificial intelligence can improve the quality of education (SDG 4) (Ma et al., 2024; Abbasi et al., 2025), optimize medical services (SDG 3) (Webster, 2023; Yip, 2024), create new types of employment (Acemoglu and Restrepo, 2018a; Damioli et al., 2024), and enhance the fairness of income distribution (SDG 10) (Chen et al., 2024; Du et al., 2024). In the environmental dimension, artificial intelligence can facilitate the transition to clean energy (SDG 7) (Chen et al., 2021; Ding C. et al., 2024) and improve agricultural production efficiency and sustainability (SDG 2) (Hu and You, 2024; Hoseinzadeh et al., 2024). However, the development of artificial intelligence is accompanied by numerous challenges. The widespread application of automation technology may replace low-skilled labor, impacting the job market (Frey and Osborne, 2017; Acemoglu and Restrepo, 2020). Meanwhile, advances in artificial intelligence technology may also exacerbate income inequality (Xin and Ye, 2024; Zhang et al., 2024). In addition, without effective regulation, experimental AI applications introduced by large technology firms in areas such as inclusive finance, due to issues including algorithmic bias, lack of transparency, and data misuse, could amplify social inequities and pose risks to sustainable development in low- and middle-income countries (Truby, 2020).

Artificial intelligence not only exerts direct effects but also enhances its own and related factors’ potential through integration and synergy with other elements, thereby generating more effective impacts across multiple domains of sustainable development. Recent studies have increasingly examined the combined effects of artificial intelligence and other factors on sustainable development. For instance, the integration of artificial intelligence with citizen science can create synergies that improve the monitoring and advancement of the Sustainable Development Goals (Fraisl et al., 2025). In the financial sector, artificial intelligence enhances the accessibility and efficiency of inclusive financial services (Fazal et al., 2025), offering new pathways to achieve related SDGs. Rasheed and Yuhuan (2025) argue that the joint influence of artificial intelligence, the circular economy, and energy intensity yields stronger effects on climate action (SDG 13), highlighting the importance of cross-domain technological synergy. Ahmed et al. (2025) explore the dynamic relationship between artificial intelligence and renewable energy sources such as solar power, revealing their potential to jointly advance sustainable development. Furthermore, Rash et al. (2025) found that the interaction between geopolitical risks and semiconductor-related factors may weaken their respective positive effects on renewable energy production.

However, existing studies have largely focused on the impact of artificial intelligence on a single dimension of economic, social, or environmental, and few have examined its role in the achievement of overall SDGs from a holistic perspective. Although a limited number of studies have attempted to evaluate the impact of artificial intelligence on multiple or even all 17 SDGs, they mainly rely on expert interviews (Vinuesa et al., 2020; Hannan et al., 2021), literature reviews (Bachmann et al., 2022), or the Delphi method (Ametepey et al., 2024), focusing on qualitative analysis. Costa et al. (2025) attempted to capture the nuanced contributions of artificial intelligence patents to individual SDGs by calculating the cosine similarity between keywords in AI patent texts and SDG-related keywords. Spagnuolo et al. (2025) employed panel data models to investigate the heterogeneous effects of artificial intelligence across the 17 SDGs. However, their constructed SDG indicators system has notable limitations, their sample is restricted to European state-owned enterprises, and they do not thoroughly explore the underlying mechanisms or heterogeneity in AI’s impact.

Overall, existing studies mainly focus on the impact of artificial intelligence on individual SDGs, lacking a systematic perspective to examine artificial intelligence’s impact and pathway on achieving the overall SDGs. Although a few studies have explored the relationship between artificial intelligence and SDGs at multi-dimensional or overall levels, they remain primarily qualitative and lack systematic quantitative evidence as well as in-depth identification of the pathways through which AI exerts its influence. However, such a research topic is urgently needed because the advancement of the SDGs faces numerous challenges. Artificial intelligence can not only directly affect the achievement of each SDG, but may also indirectly drive the achievement of other goals by acting on certain targets, and indirectly influence the overall SDGs through various pathways. Therefore, systematic examination of impact and pathway of artificial intelligence on achieving SDGs in China is of great practical significance for accelerating sustainable development in China and other countries. The main contributions of this study are as follows: First, it establishes an SDGs indicator system suitable for China’s provincial context, supporting the quantitative assessment of regional sustainable development. Second, drawing on a large-scale panel dataset, we offer the first quantitative analysis of the impact of artificial intelligence on the achievement of overall SDGs in China, along with its underlying mechanisms, while also examining regional heterogeneity and spatial dependence, thereby addressing methodological and empirical gaps in the existing literature.

In order to make up for shortcomings of existing studies, this study constructs a localized SDGs evaluation index system based on the panel data of 30 provinces (autonomous regions and municipalities) in China from 2010 to 2023. Applying the entropy method, two-way fixed effect model, mediating effect model, and spatial error model, it systematically explores the impact of artificial intelligence on the achievement of SDGs, including its magnitude, mechanisms, and regional spillover effect. This study focuses on the following questions: (1) To what extent does artificial intelligence affect the achievement of SDGs in China‘s provinces? (2) What is the specific transmission pathway? (3) Does the impact vary across regions? (4) Is there spatial dependence? Based on multi-source indicator data, a stepwise modeling approach is employed to examine the direct effect, mediating mechanisms, and spatial association characteristics of artificial intelligence on the achievement of SDGs. The overall analytical framework is shown in Figure 1.

Figure 1

Figure 1. Analytical framework.

2 Theoretical foundation and research hypotheses

2.1 Sustainable development theory

Sustainable development theory was introduced by the World Commission on Environment and Development (Keeble, 1988) in its report Our Common Future, defining sustainable development as “development that meets the needs of the present without compromising the ability of future generations to meet their own needs.” The theory emphasizes the coordination and balance among three dimensions of economy, society and environment. The United Nations’ adoption of the 17 SDGs in 2015 translated this concept into an actionable global framework.

Traditional industrialization models, which rely heavily on resource-intensive industries, have driven short-term economic growth but simultaneously led to resource depletion, ecological degradation, and social inequality, thereby constraining improvements in development quality. Consequently, achieving sustainable development hinges on transforming development pathways toward greater inclusiveness, environmental sustainability, and efficiency (Mentes, 2023), with technological applications playing a critical role in this transition (Singh et al., 2024). As a widely applied technology, artificial intelligence has increasingly been integrated into production, governance, and service delivery. Its applications in areas such as energy management, environmental monitoring, and public services influence the SDGs framework by optimizing resource allocation and enhancing cross-sectoral coordination. Drawing on the holistic perspective of sustainable development theory, this study proposes Hypothesis 1.

Hypothesis 1

Hypothesis 1. The level of artificial intelligence development has a positive effect on the overall achievement of the SDGs.

2.2 Endogenous growth theory

Traditional neoclassical growth theory treats technological progress as exogenous to the economic system, positing that long-run economic growth will eventually stagnate under the constraint of diminishing returns to factors of production. Endogenous growth theory, systematically advanced in the late 1980s by Romer (1990), Lucas (1988), and others, overcomes this limitation by emphasizing that sustained economic growth originates from internal drivers within the economic system, particularly knowledge accumulation, research and development (R&D) investment, and human capital formation. Central to this theory is the introduction of assumptions of increasing returns and imperfect competition. It highlights that knowledge is non-rivalrous and partially excludable, enabling innovations by one economic agent to be adopted by others at low marginal cost, thereby generating positive knowledge spillovers that continuously enhance economy-wide productivity.

Within this framework, technological innovation is not an exogenous shock but rather the outcome of endogenous choices made by economic agents in response to market incentives, institutional environments, and policy support. The development of artificial intelligence relies on continuous algorithmic iteration, data inputs, and computational power upgrades, making it inherently a knowledge-intensive activity. Its R&D process itself constitutes a core component of technological innovation (Li X. et al., 2025). More importantly, artificial intelligence applications often drive broader technological progress by fostering new technologies, improving existing processes, or enabling novel service models (Behera et al., 2025). For example, artificial intelligence accelerates the advancement of smart grid technologies in the energy sector, promotes automation and flexible manufacturing systems in industry, and facilitates the diffusion of precision agriculture techniques in rural areas. These AI-driven innovations enhance resource efficiency, mitigate environmental externalities, and improve access to public services, thereby exerting substantive impacts on the achievement of SDGs. Based on this, we propose Hypothesis 2.

Hypothesis 2

Hypothesis 2. Technological innovation mediates the relationship between artificial intelligence and the achievement of SDGs.

2.3 Techno-economic paradigm theory

Major technological transformations drive the transformation of production methods while simultaneously promoting economic structural optimization and institutional change. The techno-economic paradigm theory was proposed by Perez (2002) to explain how successive technological revolutions trigger systemic adjustments in industrial organization, labor skill structures, and institutional arrangements through the diffusion of dominant technology clusters. The theory posits that each technological wave, such as those driven by the steam engine, electricity, or information technology, gives rise to a techno-economic paradigm with broad penetration effects. Such paradigms not only enhance overall economic efficiency but also reconfigure the division of labor and the foundations of societal capabilities.

Artificial intelligence, as a general-purpose technology, is currently driving the evolution of a new techno-economic paradigm. Its applications are reshaping the skill demand structure in labor markets (Wang and Jiao, 2025). On one hand, artificial intelligence substitutes repetitive and routine physical and cognitive tasks through automation, reducing demand for low-skilled labor. On the other hand, it creates new demand for advanced cognitive skills in areas such as data analytics, intelligent decision-making, and human-AI collaboration. This structural shift encourages individuals to increase their investment in education and training, thereby contributing to the enhancement of regional human capital (Guarascio and Reljic, 2025). In addition, artificial intelligence applications in public services, such as intelligent tutoring systems that support personalized learning and AI-assisted diagnostics that strengthen primary healthcare capacity, have expanded access to high-quality education and health services to some extent. This in turn promotes the accumulation and more balanced development of human capital. Drawing on techno-economic paradigm theory, artificial intelligence fosters inclusive development, a key component of the SDGs, by transforming skill structures and reshaping the delivery of public services. Therefore, this study proposes Hypothesis 3.

Hypothesis 3

Hypothesis 3. Human capital mediates the relationship between artificial intelligence and the achievement of SDGs.

Based on the above theoretical foundation and research hypotheses, this paper presents a conceptual diagram (see Figure 2) that illustrates the relationships among all variables in this study, connects them with relevant theoretical constructs, and delineates the pathway for the formation of the research hypotheses.

Figure 2

Figure 2. Conceptual framework diagram. Where, H1, H2, and H3 represent Hypotheses 1, 2, and 3 proposed in this study, respectively.

3 Models and variables

3.1 Variable selection

3.1.1 Dependent variable

This paper takes China’s SDGs as the dependent variable. The United Nations has released the Sustainable Development Goals system which contains 17 goals and 169 specific indicators. In view of China’s unique national conditions and problems existing in data acquisition, this paper refers to studies of Xu et al. (2020), Xing et al. (2024), and Li C. G. et al. (2025), and according to the global Sustainable Development Goal indicators framework released by the United Nations (Pintér et al., 2017) and the China SDG Indicators Construction and Progress Assessment Report 2018 released by the Chinese Academy of Environmental Planning (CAEP), this paper has finally selected 121 sub-targets suitable for provincial level assessment and constructed China’s SDG indicators system (for details of the specific indicators, see Supplementary Table S1 of Supplementary Materia1 S1).

In this paper, the entropy method is used to calculate the score of each of the 17 SDGs, and on this basis, further computes the total score of the overall SDGs, in order to measure the achievement level of each individual SDG and the overall SDGs. To distinguish the progress of the overall SDGs from 17 individual goals, we refer to the latter as sub-SDGs. The entropy method assigns weights objectively according to the amount of information provided by each indicator, and indicators that are more informative are assigned greater weights. The specific calculation steps of the entropy method are as follows:

1. Standardization of indicator data.

Positive indicator:

$X_{ij}^{\prime }=\frac{X_{ij}-\min \left(X_{ij}\right)}{\max \left(X_{ij}\right)-\min \left(X_{ij}\right)}(1)$

Negative indicator:

$X_{ij}^{\prime }=\frac{\max \left(X_{ij}\right)-X_{ij}}{\max \left(X_{ij}\right)-\min \left(X_{ij}\right)}(2)$

where, $X_{ij}$ and $X_{ij}^{\prime }$ respectively represent the original value and standardized value of indicator $j$ for province $i$, $\max \left(X_{ij}\right)$ and $\min \left(X_{ij}\right)$ respectively denote the maximum and minimum value of indicator $j$ for province $i$.

2. Calculate the proportion of indicator $j$ for province $i$ ($P_{ij}$)

$P_{ij}=\frac{X_{ij}}{{\displaystyle {\sum }_{i=1}^n}{X}_{ij}}(3)$

3. Calculate the information entropy of indicator $j$ ($e_j$)

$e_j=-\frac{1}{\mathit{Ln}\left(n\right)}{\displaystyle \sum _{i=1}^n}{P}_{ij}\times \mathit{Ln}\left(P_{ij}\right)\left(0\le e_j\le 1\right)(4)$

4. Calculate the coefficient of variation of indicator $j$ ($a_j$)

$a_j=1-e_j(5)$

5. Calculate the weight of indicator $j$ ($w_j$)

$w_j=\frac{d_j}{{\displaystyle {\sum }_{j=1}^m}{d}_j}(6)$

6. Calculate the comprehensive score of the individual indicator for province $i$ ($SD{G}_{si}$).

$SD{G}_{si}={\displaystyle \sum _{j=1}^m}{w}_j{X}_{ij}(7)$

In Equations 1–7, $i\left(i=1,2,...,n\right)$ represents a province, $j\left(j=1,2,...,m\right)$ denotes each SDG indicator, $n$ stands for the number of provinces, and $m$ indicates the number of indicators.

3.1.2 Core independent variable

In this paper, the core independent variable is artificial intelligence (AI). The deployment scale of industrial robots is an important manifestation of AI applications, directly reflecting the maturity and cost-effectiveness of AI algorithms (Cockburn et al., 2018). An industrial robot is a reprogrammable and automatically controlled multi-purpose machine. By virtue of computer vision and deep learning technologies, it can perform functions such as image recognition, real-time control, and prediction, providing efficient solutions for industrial applications (Zhang and Yan, 2021). The greater the number of industrial robots, the higher the application level of AI in this field. Therefore, with reference to the research methods of Du and Lin (2022) and Li et al. (2022), this paper adopts the installation density of industrial robots to measure the investment level in AI and performs a logarithmic transformation on this variable. Specifically, based on the installed quantity of industrial robots in various industries in China published by the International Federation of Robotics (IFR), combined with the proportion of each province’s employment in the national total from China’s Labor Statistical Yearbook, the installation density of industrial robots in each province of China is calculated by multiplying the proportion of employment in each province by the installed quantity of robots in each industry of China.

3.1.3 Mediating variables

1. Technological innovation. Artificial intelligence aims to promote technological innovation by accelerating knowledge generation, enhancing technology spillovers, improving learning capacities, and accumulating elements of innovation. Technological innovation further promotes productivity improvement, industrial structure optimization, sustainable economic growth, and social progress. Referring to Liu et al. (2020), this paper uses the number of invention patents granted in each province to measure technological innovation.

2. Human capital. Artificial intelligence changes the structure of labor markets, enables workers to improve their skills, and provides efficient and personalized means for human capital development. High-quality human capital can better adapt to and promote technological innovation, and contribute to sustainable economic and social development. Referring to Ran et al. (2023), this paper measures human capital by the proportion of higher education students in each province’s total population.

3.1.4 Control variables

Referring to research methods of Lu et al. (2024), Zhou J. P. et al. (2024), and Škare et al. (2024), this study selects the level of economic development, the level of financial development, the degree of industrialization, the degree of trade openness, population density and environmental regulation as control variables. The specific connotation of all variables is shown in Table 1.

Table 1

Table 1. The connotation of all variables.

3.2 Data source

Due to the limitation of data availability (excluding Xizang, Hong Kong, Macao and Taiwan), this study selects the panel data of 30 provinces (autonomous regions and municipalities) in China from 2010 to 2023. The data are sourced from the China Statistical Yearbook, China Statistical Yearbook on Energy, China Statistical Yearbook on Environment, China Statistical Yearbook on Fishery, China Statistical Yearbook on Science and Technology, China Fiscal Yearbook, statistical yearbooks of all provinces, and database of the National Bureau of Statistics. For missing values in certain variables (with at most three consecutive years missing), we adopt the linear interpolation method to fill the gaps, following Che et al. (2024). The descriptive statistical result of each variable is detailed in Supplementary Table S1 of Supplementary Material S2.

3.3 Model setting

3.3.1 Two-way fixed effect model

In order to systematically examine the impact of artificial intelligence on the achievement of SDGs in China, this study adopts two-way fixed effect model for the main regression analysis, drawing on research methods of Aust et al. (2020) and Yang et al. (2023). This model is widely used in panel data analysis to control for time-invariant regional heterogeneity and region-invariant time effect, thereby mitigating potential estimation bias caused by omitted variables. The model is constructed as follows:

$CSSDG{s}_{it}={\alpha }_0+{\alpha }_{1}LnA{I}_{it}+\gamma Contro{l}_{it}+{\mu }_i+{\eta }_t+{\epsilon }_{it}(8)$

$CSSDG{s}_{it}={\alpha }_0+{\alpha }_{1}L.LnA{I}_{it}+\gamma Contro{l}_{it}+{\mu }_i+{\eta }_t+{\epsilon }_{it}(9)$

$CSSDG{s}_{it}={\alpha }_0+{\alpha }_{1}L2.LnA{I}_{it}+\gamma Contro{l}_{it}+{\mu }_i+{\eta }_t+{\epsilon }_{it}(10)$

$SDG{j}_{it}={\alpha }_0+{\alpha }_{3}LnA{I}_{it}+\gamma Contro{l}_{it}+{\mu }_i+{\eta }_t+{\epsilon }_{it}(11)$

In Equations 8–11, $CSSDG{s}_{it}$ represents the total score of the 17 SDGs for province $i$ in year $t$. $SDG{j}_{it}$ represents the score of each sub-SDG for province $i$ in year $t$. $LnA{I}_{it}$ represents artificial intelligence level for province $i$ in year $t$. $L.LnA{I}_{it}$ represents artificial intelligence’s one-period-lagged effect for province $i$ in year $t$. $L2.LnA{I}_{it}$ represents artificial intelligence’s two-period-lagged effect for province $i$ in year $t$. $Contro{l}_{it}$ is a series of control variables. ${\alpha }_{1\sim 3}$ and $\gamma$ are coefficient to be estimated, and ${\alpha }_0$ is a constant term. ${\mu }_i$ represents the provincial individual fixed effect, ${\eta }_t$ represents the time fixed effect, ${\epsilon }_{it}$ represents the random error disturbance term.

3.3.2 Mediation effect model

In order to explore whether artificial intelligence affects the achievement of SDGs by promoting technological innovation and human capital, this paper adopts a mediation effect model for testing. Considering that the mechanisms of action of artificial intelligence may differ across fields, the onset time and intensity of its effect may also vary: some pathways may exhibit a relatively rapid response, while others need to undergo a certain adjustment process before gradually emerging. A relatively close relationship exists between artificial intelligence (AI) and technological innovation (TI). Artificial intelligence can directly promote technological innovation activities such as patent output in the current period by optimizing R&D processes and improving information processing efficiency (Liu et al., 2025), demonstrating a certain immediate effect. Therefore, when testing this mediating mechanism, this paper uses the one-period lagged values of artificial intelligence and technological innovation to examine their impact on the current progress of SDGs, with the transmission path specified as AI_t−1 → TI_t−1 → SDGs_t.

For human capital, the “task model” proposed by Acemoglu and Restrepo (2018b) suggests that technological change alters skill demand structures, but worker retraining, curriculum adjustments in education systems, and institutional responses require substantial time. Therefore, the impact of artificial intelligence on human capital structure is inherently lagged, with the transformation process typically unfolding through education reform, skills training, and generational turnover. However, given that education is a core component of the SDGs, improvements in human capital can have an immediate positive effect on the achievement of SDGs once realized. Based on this reasoning, to test the mediating role of human capital, this study employs the one-period lagged value of artificial intelligence as the explanatory variable and current-period human capital as the mediator to assess its effect on the current progress of SDGs, specifying the pathway as AI_t−1 → HC_t → SDGs_t.

The mediation model is specified as follows:

$LnT{I}_{it}={\beta }_0+{\beta }_{1}LnA{I}_{it}+\gamma Contro{l}_{it}+{\mu }_i+{\eta }_t+{\epsilon }_{it}(12)$

$CSSDG{s}_{it}={\theta }_0+{\theta }_{1}L.LnA{I}_{it}+\delta L.LnT{I}_{it}+\gamma Contro{l}_{it}+{\mu }_i+{\eta }_t+{\epsilon }_{it}(13)$

$H{C}_{it}={\beta }_0+{\beta }_{2}L.LnA{I}_{it}+\gamma Contro{l}_{it}+{\mu }_i+{\eta }_t+{\epsilon }_{it}(14)$

$CSSDG{s}_{it}={\theta }_0+{\theta }_{2}L.LnA{I}_{it}+\rho H{C}_{it}+\gamma Contro{l}_{it}+{\mu }_i+{\eta }_t+{\epsilon }_{it}(15)$

In Equations 12–15, variables $LnTI$ and $HC$ respectively represent technological innovation and human capital, and other variables are set in the same way as in Equation 8. ${\theta }_1$ and ${\theta }_2$ both reflect direct impact of artificial intelligence on achieving SDGs, and coefficient products ${\beta }_1\times \delta$ and ${\beta }_2\times \rho$ respectively measure the mediation effect of technological innovation and human capital.

4 Results

4.1 Spatiotemporal evolution of China’s SDGs achievement

The progress of China’s SDGs has gradually strengthened in space and time (Figures 3a–d). From 2010 to 2014, before formal introduction of SDGs, China’s overall sustainable development level was relatively low, and only eastern coastal provinces and a few central provinces reached medium development level (Figures 3a,b). Regional differences in the level of sustainable development during this period are significant, reflecting imbalance of China’s regional development at that time. Since formal introduction of the United Nations Sustainable Development Goals in 2015, the Chinese government has attached great importance to and actively promoted the implementation of SDGs. From 2015 to 2023, China’s overall sustainable development level has been significantly improved, with the development trend gradually expanding from eastern coastal areas to inland areas. Central and western regions have also made significant progress (see Figures 3c,d).

Figure 3

Figure 3. Temporal and spatial evolution of the progress of SDGs in Chinese provinces. (a–d) are comprehensive SDGs scores of each province in 2010, 2014, 2019, and 2023, respectively. NA indicates that the data is not available.

4.2 The impact of artificial intelligence on SDGs achievement

Artificial intelligence development plays a significant role in promoting the achievement of SDGs in China (Figure 4; Supplementary Table S2 in Supplementary Material S2). The benchmark regression analysis shows that the impact coefficient of artificial intelligence on the achievement of overall SDGs is 0.191, significantly positive at the 1% level, indicating that it has direct positive impact on achieving overall SDGs. Further analysis shows that one-period lag and two-period lag effect of artificial intelligence are also significantly positive at the 1% level, and the promotion effect of two-period lag is stronger than that of current period. This shows that the promoting effect of artificial intelligence on achieving overall SDGs is not only apparent in the current period, but also has continuous and cumulative effect which become more significant over time, providing more lasting driving forces for sustainable development.

Figure 4

Figure 4. The impact of artificial intelligence on the achievement of overall SDGs. The numerical values above the bars represent the estimated regression coefficient, and the symbols *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively (p < 0.1, p < 0.05, p < 0.01). The vertical lines (error bars) above the bars show the 90% confidence intervals of the coefficient. LnAI, L.LnAI, and L2.LnAI represent the current value, one-period lagged value, and two-period lagged value of artificial intelligence, respectively. The figure shows the impact coefficient of these variables on achieving overall SDGs.

Artificial intelligence contributes significantly to the achievement of some sub-SDGs (Figure 5; Supplementary Table S3 in Supplementary Material S2). Specifically, the impact coefficient of artificial intelligence on the progress of SDG1(No Poverty), SDG4 (Quality Education), SDG6 (Clean Water and Sanitation), SDG8 (Decent Work and Economic Growth), SDG9 (Industry, Innovation and Infrastructure), SDG12 (Responsible Consumption and Production) and SDG13 (Climate Action) are all statistically significantly positive, indicating that the development of artificial intelligence effectively promotes the progress of these goals, but there are differences in the degree of its promotion. Among them, economic development-related goals (SDG8 and SDG9) and climate action (SDG13) are promoted most effectively.

Figure 5

Figure 5. The impact of artificial intelligence on the achievement of sub-SDGs. The numerical values on the left and right sides of the bars represent the estimated regression coefficient, and the symbols *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively (p < 0.1, p < 0.05, p < 0.01). The vertical lines (error bars) on the bars show the 90% confidence intervals of the coefficient. Sub-SDGs refer to the 17 specific goals from SDG 1 to SDG 17. The figure shows the impact coefficient of artificial intelligence on achieving each SDG (from SDG 1 to SDG 17). Specifically, SDG2_1 represents the estimated effect of artificial intelligence on SDG2 after controlling for rural agricultural mechanization and urbanization rate. SDG2_2 represents the estimated effect of artificial intelligence on SDG2 after further controlling for industrial structure variables (including the shares of primary, secondary, and tertiary industries in GDP).

However, the impact of artificial intelligence on SDG2 (Zero Hunger) exhibits a statistically significant negative coefficient (−0.293, p < 0.01), suggesting that artificial intelligence development may pose certain obstacles to achieving food security. To test the robustness of this result, we conduct three supplementary analyses. First, after controlling for rural agricultural mechanization rates and urban-rural population migration (corresponding to SDG2_1 in Figure 5; Supplementary Table S4 in Supplementary Material S2), the coefficient of artificial intelligence on SDG2 remains negative (−0.284, p < 0.01), though its absolute value decreases slightly. Second, after further including industrial structure variables (i.e., shares of primary, secondary, and tertiary industries in GDP) (SDG2_2), the coefficient weakens further to −0.211 (p < 0.01), indicating that industrial restructuring mitigates the negative impact of artificial intelligence to some extent. Finally, provincial heterogeneity analysis shows that this negative relationship remains prevalent in the subsample of major agricultural provinces (such as Shandong, Jilin, Inner Mongolia, and Hebei) (see Supplementary Table S4 in Supplementary Material S2). Overall, this negative association remains relatively robust across various model specifications, potentially reflecting structural effect of artificial intelligence development on the agricultural sector. However, due to limitations in data completeness, observation period length, and potential omitted variables, the causal interpretation of this finding should be made with caution.

4.3 Robustness analysis of the impact of artificial intelligence on SDGs achievement

4.3.1 Robustness test

In order to verify the robustness of the empirical analysis result, this study uses a variety of methods to systematically test the benchmark regression result.

First, replace the measure of the core independent variable (see “Robustness1” in Figure 6 and column (1) of Supplementary Table S6 in Supplementary Material S2). Referring to the research method of Zhang (2024), this paper replaces the “number of industrial robot installations” in the benchmark model with the “number of artificial intelligence enterprises” in each province of China as a new proxy variable to measure the development level of artificial intelligence for robustness test. The data are from the Pipixia Database organized and managed by Pipixia Team. In order to evaluate the correlation between the two indicators, we calculated the Pearson correlation coefficient between the number of artificial intelligence enterprises and the number of industrial robot installations. As shown in Supplementary Table S5 of Supplementary Material S2, the correlation coefficient between the two is 0.825, indicating that the two have a strong positive correlation. This result shows that although the focus is slightly different, the former reflects the agglomeration degree of AI industry ecology, and the latter reflects the application level of intelligent equipment, but the two are consistent in describing the development trend of regional artificial intelligence. The regression result show that after replacing the core independent variable, the direction and significance of the impact of artificial intelligence on the achievement of SDGs in China remain unchanged.

Figure 6

Figure 6. The robustness test of the impact of artificial intelligence on the achievement of SDGs. Among them, the values above the bar are the coefficient of the regression estimate, the symbols *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively (p < 0.1, p < 0.05, p < 0.01). The vertical line on the bar represents a 90% confidence interval of the coefficient (error bar). “Robustness1” represents the result of regression of SDGs achievement using alternative independent variable (logarithm of the number of artificial intelligence enterprises). “Robustness2” represents the regression result of the effect of artificial intelligence when the SDGs score calculated by the equal weight method is used as an alternative dependent variable. “Robustness3” reports the regression result of the impact of artificial intelligence on the achievement of SDGs after adding additional control variables such as industrial structure upgrading, urbanization rate and urban green space area on the basis of the original model. “Robustness4” shows the result of regression analysis after excluding samples from 2020 to 2021 (i.e., using data from 2013–2019 to 2022–2023). “Endogeneity1” reported the estimation result based on the system GMM method. “Endogeneity2” presents the second-stage regression result based on two-stage least squares (2SLS) with fiber cable density as the instrumental variable.

Second, adjust the measurement method of the dependent variable (see “Robustness2” in Figure 6 and Column (2) of Supplementary Table S6 in Supplementary Material S2). In the benchmark regression, the SDGs scores are weighted by the entropy method to reflect the actual contribution differences of each secondary index in the implementation process. However, the entropy method may lead to weight offset due to data distribution characteristics, thus affecting the comparability of result. Therefore, this paper draws on the practice of Arrigoni et al. (2022), and uses the equal weight method as an alternative to test the robustness. Specifically, based on the original standardized data before the entropy method, all secondary indicators under each SDG are given the same weight, and the individual scores of each SDG are calculated by weighted average. On this basis, the composite score of 17 SDGs is further summarized. The regression result show that after using the equal weight method, the influence coefficient of artificial intelligence on the achievement of SDGs is 0.080, which is significantly positive at the 1% level, which is consistent with the influence direction and significance of the benchmark result.

Third, in order to alleviate the possible bias of missing variables in the benchmark model, this paper further introduces variables such as industrial structure upgrading (measured by the ratio of tertiary to secondary industry value-added), urbanization rate (the proportion of urban population to the total population) and urban green space area (square kilometers) as additional control variables (see “Robustness3” in Figure 6 and Column (3) of Supplementary Table S6 in Supplementary Material S2), to more fully control the potential impact of regional development characteristics on the achievement of SDGs. The regression result show that the coefficient of artificial intelligence on the achievement of SDGs is still positive and statistically significant.

Fourth, eliminate the special year samples to exclude the interference of external shocks (see “Robustness4” in Figure 6 and Column (4) of Supplementary Table S6 in Supplementary Material S2). Since 2020, the COVID-19 pandemic has rapidly spread globally, severely disrupting normal production and living order in various countries and posing a significant challenge to the achievement of global SDGs (Elsamadony et al., 2022). In order to exclude the impact of COVID-19 on achieving SDGs, this study follows the methodology of Qian et al. (2024) and excludes the observational data from 2020 to 2021 during the COVID-19 pandemic. The result show that after excluding the impact of COVID-19, positive impact of artificial intelligence on achieving SDGs remains statistically significant.

4.3.2 Endogeneity test

In order to address potential endogeneity issues in the baseline regression, this study employs the system generalized method of moments (System-GMM) and the instrumental variable approach to enhance the reliability of the estimation result.

First, referring to the research design of Zhao et al. (2024), the system generalized moment estimation (System-GMM) is used to re-estimate the model (see “Endogeneity1” in Figure 6 and Column (1) of Supplementary Table S7 in Supplementary Material S2). This method is suitable for the dynamic panel model with the lag term of the explained variable, which can effectively alleviate the endogenous bias caused by the dynamic adjustment process. The result show that the estimated coefficient of artificial intelligence is 0.170, which is significantly positive at the 5% level, indicating that artificial intelligence still has a continuous positive impact on SDGs after considering the dynamics of the achievement of SDGs.

Second, in order to further identify the causal relationship, this paper draws on the practice of Li L. B. et al. (2025) and uses the two-stage least squares method (2SLS) to estimate the instrumental variables (see “Endogeneity2” in Figure 6 and Columns (2) and (3) of Supplementary Table S7 in Supplementary Material S2). Referring to Yang et al. (2025), fiber cable density (FCD) is selected as an instrumental variable of artificial intelligence, which is measured by the length of long-distance fiber cable lines per kilometer in each province. Its rationality lies in: as a digital infrastructure, the density of optical cable directly affects the access and application of artificial intelligence technology and meets the relevant requirements. At the same time, its layout is mainly affected by historical communication planning and geographical conditions, and does not directly affect the implementation of SDGs, which satisfies the exogenous hypothesis. The first stage F statistic is 27.41, which is much larger than the 10% critical value (16.38), indicating that there is no weak instrumental variable problem. The result of the second stage regression show that the estimation coefficient of artificial intelligence on the achievement of SDGs is still significantly positive, and the magnitude is improved. It is worth noting that although fiber cable density satisfies the relevance and exogeneity assumptions as an instrumental variable, its exclusion restriction remains subject to potential limitations. The improvement of optical cable infrastructure may indirectly affect the achievement of SDGs through alternative ways, such as enhancing information transmission efficiency and increasing access to digital services (Li M. et al., 2025). As these indirect channels cannot be entirely ruled out, instrumental variable estimates should be interpreted with caution, reflecting a strong association between artificial intelligence and SDG progress rather than a strictly causal effect.

In summary, regardless of variable substitution, measurement method adjustment, model expansion or endogeneity test, the positive impact of artificial intelligence on the achievement of SDGs remains statistically significant under various estimation methods, and the consistency of the result further verifies the robustness of the benchmark regression conclusions.

4.4 Impact pathway of artificial intelligence on SDGs achievement

Artificial intelligence is positively correlated with technological innovation and human capital, and technological innovation and human capital are also positively correlated with the achievement of SDGs (Figure 7; Supplementary Table S11 in Supplementary Material S2), indicating that they are important transmission pathway for artificial intelligence to influence the achievement of SDGs. As shown in Figures 7b,c, the estimated coefficient of artificial intelligence for technological innovation and human capital are both positive, indicating that its development helps to promote technological progress and optimize the structure of human capital, and has a stronger role in promoting technological innovation (${\beta }_1=0.836>{\beta }_2=0.394$). At the same time, the estimated coefficient of technological innovation and human capital to SDGs are also positive, reflecting its positive role in the process of sustainable development.

Figure 7

Figure 7. The impact pathway of artificial intelligence on the achievement of SDGs. Among them, the symbols *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively (p < 0.1, p < 0.05, p < 0.01). The icons for “artificial intelligence,” “technological innovation” and “human capital” are from The Noun Project (https://theno.unpro.ject.com/), and the icons for SDGs are from The Global SDG Indicators Data Platform (https://unstats.un.org/sdgs/dataportal). (a) Is the total effect of artificial intelligence on achieving SDGs. (b,c) show the test result of technological innovation and human capital as mediating variables, respectively. The result of Sobel test and Bootstrap test both show that technological innovation and human capital play a role as intermediary transmission channels.

Further comparison shows that the direct effect coefficient of artificial intelligence on SDGs in Figures 7b,c are smaller than the total effect coefficient (${\theta }_1=0.172<{\alpha }_1=0.191$; ${\theta }_2=0.018<{\alpha }_1=0.191$) in Figure 7a, indicating that both technological innovation and human capital play a partial mediating role, with the mediating effect of technological innovation being stronger than that of human capital. Specifically, the proportion of the indirect effect through technological innovation is 9.63% ((0.022 × 0.836)/0.191% × 100%), while that through human capital is 5.98% ((0.029 × 0.394)/0.191% × 100%), suggesting that technological innovation plays a more prominent role in the pathway through which artificial intelligence influences the achievement of SDGs.

4.5 Heterogeneous effect of artificial intelligence on SDGs achievement

The impact of artificial intelligence on achieving SDGs shows obvious heterogeneity across different regions in China (Figure 8; Supplementary Tables S6–S8 in Supplementary Material S2), mainly closely related to the levels of economic development, artificial intelligence, and SDGs performance in each region. In terms of economic development level, the impact coefficient of artificial intelligence on achieving SDGs is statistically significantly positive in more economically developed eastern provinces, indicating that artificial intelligence plays an important role in promoting the achievement of SDGs and brings relatively significant marginal benefits to these regions. In contrast, the impact coefficient of artificial intelligence in central and western regions is not statistically significant.

Figure 8

Figure 8. The heterogeneous impact of artificial intelligence on the achievement of SDGs. Among them, the values above the bar are the coefficient of regression estimation, the symbols *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively (p < 0.1, p < 0.05, p < 0.01). The vertical line on the bar represents the 90% confidence interval of the coefficient (error bar). “Eastern,” “Central,” and “Western” represent China’s eastern, central and western regions, respectively. “Low-AI” and “High-AI” stand for “low artificial intelligence level group” and “high artificial intelligence level group” respectively, with the grouping based on the average value of artificial intelligence (industrial robot installation density). “Low-SDGs” and “High-SDGs” correspond to “poor SDGs performance group” and “good SDGs performance group” respectively, grouped according to the average value of the total score of overall SDGs. Based on the First National Economic Census Main Data Bulletin (No. 1) released by the National Bureau of Statistics in 2005 and the regional division by Bai et al. (2014), this paper divides China into three major economic regions: eastern, central and western (see the notes in Supplementary Table S6 for specific division). Referring to the study by Kondo et al. (2009), the average values of artificial intelligence and the total score of overall SDGs are used as thresholds for grouping to conduct heterogeneity analysis. The average value can well reflect the central tendency of data, enabling the division of data into two groups (above or below the average) to further judge differences under different levels. In addition, following the research of Lu et al. (2020), a sensitivity analysis is carried out on the grouping of artificial intelligence and the total score of overall SDGs to test changes in the impact coefficient when the artificial intelligence level changes by 10%. The result show that the impact coefficient of heterogeneous groups changes slightly under different conditions, indicating that the estimation result have a certain stability (see Supplementary Tables S7 and S8 for specific sensitivity analysis result).

In terms of artificial intelligence level, the impact coefficient of artificial intelligence on achieving SDGs is statistically significantly positive in provinces with a lower level of artificial intelligence development, indicating that the application of artificial intelligence in these regions can effectively promote the achievement of SDGs and bring significant marginal benefits. However, in provinces with a higher level of artificial intelligence development, this driving effect is insignificant, probably because artificial intelligence technology in these regions has been relatively mature, the release of technology dividends is close to saturation, and the space for further improvement is relatively limited.

In terms of SDGs performance, in provinces with better SDGs performance, the impact coefficient of artificial intelligence on achieving SDGs is statistically significantly positive, indicating that these provinces can adopt artificial intelligence technology more effectively, thereby better promoting the achievement of local SDGs. However, in provinces with poor SDGs performance, the impact coefficient of artificial intelligence on achieving SDGs is not statistically significant, which may suggest that these regions face greater challenges in technology application and resource integration, limiting the facilitating role of artificial intelligence in achieving SDGs.

4.6 Spatial spillover effect of artificial intelligence on SDGs achievement

4.6.1 Spatial autocorrelation test

In this paper, a spatial weight matrix based on geographical distance is constructed, and the weight is determined by the reciprocal of the square of the latitude and longitude distance between the two provinces. The specific calculation formula is as follows:

$W_{ij}=\frac{1}{d_{ij}^2},i\ne j(16)$

In Equation 16, $W_{ij}$ denotes the geographical distance weight between province $i$ and province $j$, and $d_{ij}$ denotes the latitude and longitude distance between the two provinces. When $i=j$ (i.e., the same province), the corresponding weight W is set to 0.

Based on this spatial weight matrix, a global Moran’s I test was conducted for artificial intelligence (AI) and the Sustainable Development Goals (SDGs) across 30 Chinese provinces (excluding Tibet, Hong Kong, Macao, and Taiwan) from 2010 to 2023 to examine whether their spatial distributions exhibit clustering patterns. As shown in Table 2, the Moran’s I indices for both AI and SDGs are consistently positive across all years, with p-values below 0.1. This indicates statistically significant positive spatial autocorrelation at the provincial level, suggesting that regions with similar levels of AI development or SDGs achievement tend to cluster geographically, reflecting underlying spatial dependence.

Table 2

Table 2. Moran’s I index and its statistical test.

4.6.2 Selection of spatial econometric model

In order to determine the appropriate spatial model form, LM test and Hausman test are carried out in this paper. As shown in Table 3, Robust LM-Error is significant at the 1% level, while Robust LM-Lag is not significant, indicating that there is a significant spatial error dependence in the data, but no obvious spatial lag effect is found, so the spatial error model (SEM) is more appropriate. In addition, the Hausman test result (p = 0.008 < 0.050) indicate that a fixed effect model should be used to control unobservable individual heterogeneity.

Table 3

Table 3. Related test result of model selection.

In this paper, the spatial error model is selected for analysis, and the model is set as follows:

$CSSDG{s}_{it}={\alpha }_0+{\alpha }_{1}LnA{I}_{it}+\gamma Contro{l}_{it}+\lambda W_{ij}{\epsilon }_{it}+{\mu }_i+{\eta }_t(17)$

In Equation 17, $W_{ij}$ is the spatial weight, $\lambda$ is the spatial autoregressive coefficient of the error term, reflecting the degree of spatial dependence, and the remaining variables and symbols are consistent with Formula 8.

4.6.3 Spatial spillover effect analysis

Table 4 reports the spatial error model (SEM) estimation result of the impact of artificial intelligence on SDGs. The result show that the coefficient of artificial intelligence is 0.116, which is significantly positive at the 5% level, indicating that its development level is positively correlated with the degree of SDGs achievement. At the same time, the spatial autoregressive coefficient $\lambda$ of the error term is 0.720, which is significant at the 1% level, indicating that the model residuals have significant spatial dependence, that is, the unobserved factors in the adjacent areas have a certain impact on the achievement of local SDGs, reflecting the implicit spatial linkage characteristics between regions in the process of sustainable development. The above result show that when studying the impact of artificial intelligence on sustainable development goals, considering spatial dependence helps to more accurately identify variable relationships and avoid estimation errors caused by ignoring spatial structure.

Table 4

Table 4. Parameter estimation result of the spatial error model.

To further explore the potential spatial spillover effect of artificial intelligence on the achievement of SDGs, Supplementary Table S12 in Supplementary Material S2 reports the decomposition result based on the spatial lag model (SAR). Although both the LM test and Wald test support the spatial error model (SEM) as the preferred specification, the SAR analysis indicates that artificial intelligence has a positive impact on local SDGs, with a positive indirect effect, suggesting the presence of regional spillover effects. This analysis serves as a supplementary reference, while the overall spatial analysis remains grounded in the SEM result.

5 Discussion

Understanding the impact and specific mechanisms of artificial intelligence on the achievement of SDGs is of great significance for optimizing policy design in the digital era and addressing the challenges of fairness and efficiency in technology diffusion (Khamis et al., 2019). Based on China’s national level data, this study constructs an SDGs index system covering economic, social and environmental dimensions, and uses the deployment density of industrial robots to measure the development level of artificial intelligence and systematically evaluate its impact on the progress of SDGs. Compared with the existing literature focusing on a single field, this paper identifies two key transmission paths of technological innovation and human capital, and introduces the spatial measurement method to examine the regional linkage effect, which makes up for the lack of systematic empirical research between artificial intelligence and sustainable development, and provides a reference analysis framework for countries in the process of digital transformation.

The study shows that artificial intelligence can significantly accelerate the progress of the overall SDGs, and this promotion has a cumulative effect of continuous enhancement, which indicates that continuous technology deployment may bring long-term development benefits. Local governments play an important role in the implementation of artificial intelligence applications, and their effectiveness depends to a certain extent on the completeness of local intelligent infrastructure and technical governance capabilities (Park, 2024; Elbehairy et al., 2025). Therefore, in the digital era, governments should take a proactive and leadership-oriented role by providing appropriate policy support and regulatory frameworks to address potential risks and challenges. In addition, artificial intelligence demonstrates significantly different impact on the progress of various individual SDGs, especially on economic growth (SDG8 and SDG9) and climate action (SDG13). This is consistent with existing evidence that artificial intelligence improves productivity and energy efficiency (Zhou W. et al., 2024; Li H. et al., 2025). In contrast, the impact of artificial intelligence on the progress of SDG2 (zero hunger) is negative. This result may reflect that the current application of artificial intelligence in the agricultural field mainly focuses on large-scale and capital-intensive scenarios, while the coverage of small farmers and marginalized groups is insufficient. In the absence of complementary institutions, such as equitable land tenure systems, market access mechanisms, and digital literacy training, the introduction of AI may exacerbate inequalities in the distribution of productive resources and thereby undermine the inclusiveness and resilience of food systems. Therefore, future policies can promote the application of artificial intelligence in human capital-intensive areas such as education (SDG4) and medical care (SDG3), future policies should also design more inclusive technology diffusion models and institutional support systems tailored to sensitive sectors like agriculture.

Artificial intelligence enabling sustainable development requires integration with the actual conditions of each region, adopting tailored and differentiated strategies to promote regional intelligence (Ding T. et al., 2024; Guo et al., 2025). Although China’s overall SDGs achievement has improved, regional development is still unbalanced. Heterogeneity analysis shows that artificial intelligence has the most obvious promoting effect on the realization of SDGs in the eastern region, while it is relatively limited in the central and western regions, which may be related to the gaps in infrastructure density, high-end talent reserve, digital skills penetration rate and industrial digital foundation in different regions. The result of spatial econometric analysis show that there is a significant spatial autocorrelation on artificial intelligence and the achievement of SDGs. The progress of a region in promoting artificial intelligence applications and achieving SDGs is not completely independent, and may be indirectly affected by policy diffusion, technology spillover, infrastructure connectivity, and factor mobility in neighboring regions. In the absence of regional coordination mechanisms, the uneven diffusion of AI technology may further widen the gap between regions in terms of development opportunities and resource access (Capraro et al., 2024). Furthermore, the widespread application of AI may lead to structural adjustments in employment, with automation replacing some low-skill jobs (Ajithkumar et al., 2023). In regions where the labor structure leans more traditional, such as parts of central and western China, this shift poses particular challenges given that retraining systems have yet to be adequately developed. Additionally, the trend toward centralized data resources could undermine the data sovereignty of grassroots governments, and increase the vulnerability of disadvantaged groups in algorithmic decision-making processes. Therefore, policy formulation should not focus solely on technology dissemination but also consider regional adaptability, social inclusiveness, and governance capacity building comprehensively.

Enhancing technological innovation capabilities and fostering the mobility of human capital are crucial for accelerating the achievement of SDGs (Guo et al., 2022; Ai et al., 2025). The study result show that technological innovation and human capital are two important pathways through which artificial intelligence influences the achievement of SDGs, with technological innovation having a greater contribution. By promoting technological innovation, artificial intelligence has significantly improved technical content and productivity in various fields, thus contributing strongly to the achievement of SDGs. At the same time, artificial intelligence also provides talent support for the achievement of SDGs by optimizing the structure of human capital (Lee et al., 2025) and improving the skill level and innovation ability of the labor force. Therefore, future policy efforts should prioritize strengthening investment in basic scientific research (Sinha et al., 2020), improving digital education systems (Ni et al., 2023), and facilitating cross-regional labor mobility.

This study has the following limitations: First, due to the constraint of data availability, the selection of SDG indicators is limited to 121 items, which may restrict the comprehensiveness of the analysis. Second, the application of artificial intelligence is not limited to manufacturing but also extensively involves the Internet of Things (IoT), big data analytics, and service sectors. Therefore, using only industrial robots to measure the development level of artificial intelligence may underestimate its impact on SDGs. Third, this study only explores indirect impact of artificial intelligence on achieving SDGs through technological innovation and optimization of human capital structure, without fully exploring other potential pathways, such as industrial structure upgrading, intensity of R&D investment, and the level of informatization. Future research should optimize research methodology, extend SDG indicators system, consider artificial intelligence measurement indicators from multiple aspects, and further explore other impact pathways of artificial intelligence on achieving SDGs.

6 Conclusion and policy implications

This study utilizes panel data from 2010 to 2023 for 30 provinces (autonomous regions and municipalities) in China, employing the entropy method to calculate evaluation index of SDG indicators. A two-way fixed effect model and a mediation effect model are employed to explore the impact of artificial intelligence on the achievement of SDGs and its underlying pathways, while the spatial error model (SEM) is used to preliminarily assess the spatial association between artificial intelligence and SDGs. The key findings are as follows: First, China’s overall level of sustainable development is on the rise, especially after 2015. The development trend has gradually expanded from coastal areas to inland, but regional differences are still significant. Second, artificial intelligence has significantly contributed to the achievement of China’s overall SDGs. This result is robust, and its role has a continuous and cumulative effect. However, there are differences in artificial intelligence’s contribution to the achievement of different sub-SDGs, with the most significant contributions to economic development (SDG8 and SDG9) and climate action (SDG13). Third, there is heterogeneity in the impact of artificial intelligence on achieving SDGs across different regions of China, with a more significant driving effect and greater marginal benefits in eastern regions, regions with better SDGs performance, and regions where artificial intelligence is in the early stages of development. Fourth, artificial intelligence indirectly promotes the achievement of overall SDGs by promoting technological innovation and optimizing human capital structure. Fifth, both artificial intelligence and SDGs have significant positive spatial agglomeration, and the unobserved factors in neighboring areas have obvious spillover effect on the achievement of local SDGs, indicating that there is a potential spatial dependence between regions.

Based on the above findings, we propose the following policy implications:

First, differentiated artificial intelligence application strategies should be formulated in combination with regional development basis differences. The eastern region can deepen the integration and application of artificial intelligence in key industries such as high-end manufacturing and green energy, and support capacity-building in the central and western regions through technology output and joint research and development. The central and western regions need to give priority to the layout of intelligent infrastructure, rely on grass-roots communities to provide basic digital skills training for residents, especially rural people, and improve technical accessibility. At the national level, the allocation of financial, computational and energy resources should be appropriately tilted to less developed regions to alleviate the development gap that may be exacerbated by uneven technology diffusion.

Second, take targeted support measures for the differences in the impact of different fields. In the field of agriculture, we should speed up the construction of rural intelligent agricultural machinery and Internet of things facilities, carry out operation and maintenance training for small farmers, and explore the establishment of agricultural digital transformation support mechanism. In human capital-intensive fields such as education and medical care, low-cost and easy-to-use artificial intelligence aids can be promoted to improve the accessibility and fairness of basic public services.

Third, improve the technical support and governance support. On the one hand, we will increase support for the research and development of artificial intelligence-related technologies, improve the treatment of scientific researchers, and encourage universities, enterprises and scientific research institutions to jointly promote technological transformation. At the same time, we should strengthen the training of digital skilled talents and improve the adaptability of labor force through the reform of education system and vocational training. On the other hand, the “AI-SDGs integration initiative” can be piloted locally to establish a dynamic evaluation mechanism, regularly track the actual impact of artificial intelligence applications on the social economy, and promote cross-regional data exchange and technical cooperation to enhance the linkage of infrastructure, green technology and factor flow in neighboring regions, so as to avoid further expansion of regional development gaps.

Data availability statement

The original contributions presented in the study are included in the article/Supplementary Material. Specifically, Supplementary file 3 contains the empirical analysis code used in this paper, and Table 1 includes the final dataset for all variables employed in the study. Further inquiries can be directed to the corresponding author.

Author contributions

CL: Conceptualization, Methodology, Resources, Writing – original draft. XP: Data curation, Formal Analysis, Methodology, Visualization, Writing – original draft. MY: Funding acquisition, Writing – review and editing. JG: Funding acquisition, Writing – review and editing. WW: Writing – review and editing. GZ: Writing – review and editing.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This study was supported by Guizhou Province Key Laboratory of Sovereign Blockchain (No.: ZSYS [2024]003); Innovation Exploration and Academic New Seed Project of Guizhou University of Finance and Economics (No.: 2024XSXMB15); Ministry of Education under its Academic Research Fund Tier 1 of Singapore (RG 106/24).

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. During the preparation of this study, we used Kimi AI and Doubao AI to polish the language of the manuscript, aiming to enhance its readability and linguistic quality. After using these tools, we have reviewed and revised the content as needed, and we take full responsibility for the content of the final published article.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fenvs.2025.1688466/full#supplementary-material

References

Abbasi, B. N., Wu, Y., and Luo, Z. (2025). Exploring the impact of artificial intelligence on curriculum development in global higher education institutions. Educ. Inf. Technol. 30 (1), 547–581. doi:10.1007/s10639-024-13113-z

CrossRef Full Text | Google Scholar

Acemoglu, D., and Restrepo, P. (2018a). The race between man and machine: implications of technology for growth, factor shares, and employment. Am. Economic Rev. 108 (6), 1488–1542. doi:10.1257/aer.20160696

CrossRef Full Text | Google Scholar

Acemoglu, D., and Restrepo, P. (2018b). “Artificial intelligence, automation, and work,” in The economics of artificial intelligence: an agenda (University of Chicago Press), 197–236. Available online at: https://www.nber.org/system/files/chapters/c14027/c14027.pdf.

CrossRef Full Text | Google Scholar

Acemoglu, D., and Restrepo, P. (2020). Robots and jobs: evidence from US labor markets. J. Political Economy 128 (6), 2188–2244. doi:10.1086/705716

CrossRef Full Text | Google Scholar

Ahmed, Z., Pinzon, S., and Rasheed, M. Q. (2025). Unlocking marginal SDG dynamics: a KRLS machine learning analysis of AI technologies and solar energy. Geosci. Front. 16, 102167. doi:10.1016/j.gsf.2025.102167

CrossRef Full Text | Google Scholar

Ai, H. S., Tan, X. Q., Mangla, S. K., Emrouznejad, A., Liu, F., and Song, M. L. (2025). Renewable energy transition and sustainable development: evidence from China. Energy Econ. 143, 108232. doi:10.1016/j.eneco.2025.108232

CrossRef Full Text | Google Scholar

Ajithkumar, A., David, A., Jacob, A., Alex, A., and Thomas, A. (2023). Impact of AI on employment and job opportunities. Int. J. Eng. Technol. Manag. Sci. 7 (4), 507–512. doi:10.46647/ijetms.2023.v07i04.067

CrossRef Full Text | Google Scholar

Ametepey, S. O., Aigbavboa, C., Thwala, W. D., and Addy, H. (2024). The impact of AI in sustainable development goal implementation: a Delphi study. Sustainability 16 (9), 3858. doi:10.3390/su16093858

CrossRef Full Text | Google Scholar

Arrigoni, S., Bobasu, A., and Venditti, F. (2022). Measuring financial conditions using equal weights combination. IMF Econ. Rev. 70 (4), 668–697. doi:10.1057/s41308-022-00170-y

CrossRef Full Text | Google Scholar

Aust, V., Morais, A. I., and Pinto, I. (2020). How does foreign direct investment contribute to sustainable development goals? Evidence from African countries. J. Clean. Prod. 245, 118823. doi:10.1016/j.jclepro.2019.118823

CrossRef Full Text | Google Scholar

Bachmann, N., Tripathi, S., Brunner, M., and Jodlbauer, H. (2022). The contribution of data-driven technologies in achieving the sustainable development goals. Sustainability 14 (5), 2497. doi:10.3390/su14052497

CrossRef Full Text | Google Scholar

Bai, L., Qiao, Q., Yao, Y., Guo, J., and Xie, M. H. (2014). Insights on the development progress of national demonstration eco-industrial parks in China. J. Clean. Prod. 70, 4–14. doi:10.1016/j.jclepro.2014.01.084

CrossRef Full Text | Google Scholar

Behera, B., Behera, P., Pata, U. K., Sethi, L., and Sethi, N. (2025). Artificial intelligence-driven green innovation for sustainable development: empirical insights from India's renewable energy transition. J. Environ. Manag. 389, 126285. doi:10.1016/j.jenvman.2025.126285

PubMed Abstract | CrossRef Full Text | Google Scholar

Capraro, V., Lentsch, A., Acemoglu, D., Akgun, S., Akhmedova, A., Bilancini, E., et al. (2024). The impact of generative artificial intelligence on socioeconomic inequalities and policy making. PNAS Nexus 3 (6), 191. doi:10.1093/pnasnexus/pgae191

PubMed Abstract | CrossRef Full Text | Google Scholar

Che, X. H., Zhang, H. K., Li, Z. B., Wang, Y., Sun, Q., Luo, D., et al. (2024). Linearly interpolating missing values in time series helps little for land cover classification using recurrent or attention networks. ISPRS J. Photogrammetry Remote Sens. 212, 73–95. doi:10.1016/j.isprsjprs.2024.04.021

CrossRef Full Text | Google Scholar

Chen, C., Hu, Y., Karuppiah, M., and Kumar, P. M. (2021). Artificial intelligence on economic evaluation of energy efficiency and renewable energy technologies. Sustain. Energy Technol. Assessments 47, 101358. doi:10.1016/j.seta.2021.101358

CrossRef Full Text | Google Scholar

Chen, K. M., Chen, X. Q., Wang, Z. A., and Zvarych, R. (2024). Does artificial intelligence promote common prosperity within enterprises? Evidence from Chinese-listed companies in the service industry. Technol. Forecast. Soc. Change 200, 123180. doi:10.1016/j.techfore.2023.123180

CrossRef Full Text | Google Scholar

Cockburn, I. M., Henderson, R., and Stern, S. (2018). The impact of artificial intelligence on innovation (Vol. 24449). Cambridge, MA: National Bureau of Economic Research. doi:10.7208/9780226613475-006

CrossRef Full Text | Google Scholar

Colglazier, W. (2015). Sustainable development agenda: 2030. Science 349 (6252), 1048–1050. doi:10.1126/science.aad2333

PubMed Abstract | CrossRef Full Text | Google Scholar

Costa, A., Crupi, A., Cesaroni, F., and Abbate, T. (2025). Exploring the role of artificial intelligence in addressing sustainable development. A semantic analysis of AI patents. Technovation 148, 103335. doi:10.1016/j.technovation.2025.103335

CrossRef Full Text | Google Scholar

Damioli, G., Van Roy, V., Vértesy, D., and Vivarelli, M. (2024). Drivers of employment dynamics of AI innovators. Technol. Forecast. Soc. Change 201, 123249. doi:10.1016/j.techfore.2024.123249

PubMed Abstract | CrossRef Full Text | Google Scholar

DeStefano, T., and Timmis, J. (2024). Robots and export quality. J. Dev. Econ. 168, 103248. doi:10.1016/j.jdeveco.2023.103248

CrossRef Full Text | Google Scholar

Diaz-Sarachaga, J. M., Jato-Espino, D., and Castro-Fresno, D. (2018). Is the sustainable development goals (SDG) index an adequate framework to measure the progress of the 2030 agenda? Sustain. Dev. 26 (6), 663–671. doi:10.1002/sd.1735

CrossRef Full Text | Google Scholar

Ding, C., Ke, J., Levine, M., and Zhou, N. (2024a). Potential of artificial intelligence in reducing energy and carbon emissions of commercial buildings at scale. Nat. Commun. 15 (1), 5916. doi:10.1038/s41467-024-50088-4

PubMed Abstract | CrossRef Full Text | Google Scholar

Ding, T., Li, H., Liu, L., and Feng, K. (2024b). An inquiry into the nexus between artificial intelligence and energy poverty in the light of global evidence. Energy Econ. 136, 107748. doi:10.1016/j.eneco.2024.107748

CrossRef Full Text | Google Scholar

Du, L. Z., and Lin, W. F. (2022). Does the application of industrial robots overcome the Solow paradox? Evidence from China. Technol. Soc. 68, 101932. doi:10.1016/j.techsoc.2022.101932

CrossRef Full Text | Google Scholar

Du, J. H., He, J. J., Yang, J., and Chen, X. H. (2024). How industrial robots affect labor income share in task model: evidence from Chinese A-share listed companies. Technol. Forecast. Soc. Change 208, 123655. doi:10.1016/j.techfore.2024.123655

CrossRef Full Text | Google Scholar

Dwivedi, Y. K., Hughes, L., Ismagilova, E., Aarts, G., Coombs, C., Crick, T., et al. (2021). Artificial intelligence (AI): multidisciplinary perspectives on emerging challenges, opportunities, and agenda for research, practice and policy. Int. J. Information Manage. 57, 101994. doi:10.1016/j.ijinfomgt.2019.08.002

CrossRef Full Text | Google Scholar

Elbehairy, F., Mora, L., and Soe, R. M. (2025). Who are you? Examining the multifaceted innovation roles of municipal governments in AI governance. Technovation 147, 103303. doi:10.1016/j.technovation.2025.103303

CrossRef Full Text | Google Scholar

Elsamadony, M., Fujii, M., Ryo, M., Nerini, F. F., Kakinuma, K., and Kanae, S. (2022). Preliminary quantitative assessment of the multidimensional impact of the COVID-19 pandemic on sustainable development goals. J. Cleaner Production 372, 133812. doi:10.1016/j.jclepro.2022.133812

PubMed Abstract | CrossRef Full Text | Google Scholar

Fazal, A., Ahmed, A., and Abbas, S. (2025). Importance of artificial intelligence in achieving sustainable development goals through financial inclusion. Qual. Res. Financial Mark. 17 (2), 432–452. doi:10.1108/QRFM-04-2023-0098

CrossRef Full Text | Google Scholar

Fraisl, D., See, L., Fritz, S., Haklay, M., and McCallum, I. (2025). Leveraging the collaborative power of AI and citizen science for sustainable development. Nat. Sustain. 8 (2), 125–132. doi:10.1038/s41893-024-01489-2

CrossRef Full Text | Google Scholar

Frey, C. B., and Osborne, M. A. (2017). The future of employment: how susceptible are jobs to computerisation? Technol. Forecasting Social Change 114, 254–280. doi:10.1016/j.techfore.2016.08.019

CrossRef Full Text | Google Scholar

Gohr, C., Rodríguez, G., Belomestnykh, S., Berg-Moelleken, D., Chauhan, N., Engler, J. O., et al. (2025). Artificial intelligence in sustainable development research. Nat. Sustain. 1–9. doi:10.1038/s41893-025-01598-6

CrossRef Full Text | Google Scholar

Goralski, M. A., and Tan, T. K. (2020). Artificial intelligence and sustainable development. Int. J. Manag. Educ. 18 (1), 100330. doi:10.1016/j.ijme.2019.100330

CrossRef Full Text | Google Scholar

Guarascio, D., and Reljic, J. (2025). AI and employment in Europe. Econ. Lett. 247, 112183. doi:10.1016/j.econlet.2025.112183

CrossRef Full Text | Google Scholar

Guo, H. D., Huang, L., and Liang, D. (2022). Further promotion of sustainable development goals using science, technology, and innovation. Innovation 3 (6), 100325. doi:10.1016/j.xinn.2022.100325

PubMed Abstract | CrossRef Full Text | Google Scholar

Guo, Q., Peng, Y., and Luo, K. (2025). The impact of artificial intelligence on energy environmental performance: empirical evidence from cities in China. Energy Econ. 141, 108136. doi:10.1016/j.eneco.2024.108136

CrossRef Full Text | Google Scholar

Hannan, M. A., Al-Shetwi, A. Q., Ker, P. J., Begum, R. A., Mansor, M., Rahman, S. A., et al. (2021). Impact of renewable energy utilization and artificial intelligence in achieving sustainable development goals. Energy Rep. 7, 5359–5373. doi:10.1016/j.egyr.2021.08.172

CrossRef Full Text | Google Scholar

Hoseinzadeh, S., and Garcia, D. A. (2024). Can AI predict the impact of its implementation in greenhouse farming? Renew. Sustain. Energy Rev. 197, 114423. doi:10.1016/j.rser.2024.114423

CrossRef Full Text | Google Scholar

Hu, G. Q., and You, F. Q. (2024). AI-enabled cyber-physical-biological systems for smart energy management and sustainable food production in a plant factory. Appl. Energy 356, 122334. doi:10.1016/j.apenergy.2023.122334

CrossRef Full Text | Google Scholar

Jiang, M. Y., and Ma, J. (2025). Policy, investment, and growth: how green finance drives corporate sustainable development? Int. Rev. Econ. Finance 102, 104326. doi:10.1016/j.iref.2025.104326

CrossRef Full Text | Google Scholar

Jordan, M. I., and Mitchell, T. M. (2015). Machine learning: trends, perspectives, and prospects. Science 349 (6245), 255–260. doi:10.1126/science.aaa8415

PubMed Abstract | CrossRef Full Text | Google Scholar

Keeble, B. R. (1988). The Brundtland report: ‘Our common future’. Medicine War 4 (1), 17–25. doi:10.1080/0748800880840878

CrossRef Full Text | Google Scholar

Khamis, A., Li, H., Prestes, E., and Haidegger, T. (2019). AI: a key enabler of sustainable development goals, part 1 [industry activities]. IEEE Robotics Automation Mag. 26 (3), 95–102. doi:10.1109/MRA.2019.2928738

CrossRef Full Text | Google Scholar

Kondo, N., Sembajwe, G., Kawachi, I., Van Dam, R. M., Subramanian, S. V., and Yamagata, Z. (2009). Income inequality, mortality, and self rated health: meta-analysis of multilevel studies. BMJ 339, b4471. doi:10.1136/bmj.b4471

PubMed Abstract | CrossRef Full Text | Google Scholar

Lee, C. C., Zou, J., and Chen, P. F. (2025). The impact of artificial intelligence on the energy consumption of corporations: the role of human capital. Energy Econ. 143, 108231. doi:10.1016/j.eneco.2025.108231

CrossRef Full Text | Google Scholar

Li, Y. Y., Zhang, Y. R., Pan, A., Han, M. C., and Veglianti, E. (2022). Carbon emission reduction effects of industrial robot applications: heterogeneity characteristics and influencing mechanisms. Technol. Soc. 70, 102034. doi:10.1016/j.techsoc.2022.102034

CrossRef Full Text | Google Scholar

Li, X., Li, S. Y., Cao, J. F., and Spulbar, A. C. (2025a). Does artificial intelligence improve energy efficiency? Evidence from provincial data in China. Energy Econ. 142, 108149. doi:10.1016/j.eneco.2024.108149

CrossRef Full Text | Google Scholar

Li, C. G., Chen, Z. L., Jiang, Q. T., Yue, M., Wu, L., Bao, Y. H., et al. (2025b). Impacts of government attention on achieving sustainable development goals: evidence from China. Geogr. Sustain. 6 (2), 100233. doi:10.1016/j.geosus.2024.08.011

CrossRef Full Text | Google Scholar

Li, L. B., Zhao, J. W., Yang, Y. H., and Ma, D. (2025c). Artificial intelligence and green development well-being: effects and mechanisms in China. Energy Econ. 141, 108094. doi:10.1016/j.eneco.2024.108094

CrossRef Full Text | Google Scholar

Li, M., Wang, Z., Shu, L., and Gao, H. (2025d). Broadband infrastructure and enterprise digital transformation: evidence from China. Res. Int. Bus. Finance 73, 102645. doi:10.1016/j.ribaf.2024.102645

CrossRef Full Text | Google Scholar

Li, H., Lu, Z., Zhang, Z., and Tanasescu, C. (2025e). How does artificial intelligence affect manufacturing firms' energy intensity? Energy Econ. 141, 108109. doi:10.1016/j.eneco.2024.108109

CrossRef Full Text | Google Scholar

Liu, J., Chang, H. H., Forrest, J. Y. L., and Yang, B. H. (2020). Influence of artificial intelligence on technological innovation: evidence from the panel data of China's manufacturing sectors. Technol. Forecast. Soc. Change 158, 120142. doi:10.1016/j.techfore.2020.120142

CrossRef Full Text | Google Scholar

Liu, Y. J., Shen, F. B., Guo, J., Hu, G. H., and Song, Y. G. (2025). Can artificial intelligence technology improve companies' capacity for green innovation? Evidence from listed companies in China. Energy Econ. 143, 108280. doi:10.1016/j.eneco.2025.108280

CrossRef Full Text | Google Scholar

Lu, Y. K., Sugawara, Y. M., Zhang, S., Tomata, Y., and Tsuji, I. (2020). Smoking cessation and incident dementia in elderly Japanese: the Ohsaki cohort 2006 study. Eur. J. Epidemiol. 35, 851–860. doi:10.1007/s10654-020-00612-9

PubMed Abstract | CrossRef Full Text | Google Scholar

Lu, F., Meng, J. X., and Cheng, B. D. (2024). How does improving agricultural mechanization affect the green development of agriculture? Evidence from China. J. Clean. Prod. 472, 143298. doi:10.1016/j.jclepro.2024.143298

CrossRef Full Text | Google Scholar

Lucas, R. E. Jr. (1988). On the mechanics of economic development. J. Monet. Econ. 22 (1), 3–42. doi:10.1016/0304-3932(88)90168-7

CrossRef Full Text | Google Scholar

Ma, H. L., Ismail, L., and Han, W. J. (2024). A bibliometric analysis of artificial intelligence in language teaching and learning (1990–2023): evolution, trends and future directions. Educ. Inf. Technol. 29, 25211–25235. doi:10.1007/s10639-024-12848-z

CrossRef Full Text | Google Scholar

Mentes, M. (2023). RETRACTED ARTICLE: sustainable development economy and the development of green economy in the European Union. Energy, Sustain. Soc. 13 (1), 32. doi:10.1186/s13705-023-00410-7

CrossRef Full Text | Google Scholar

Merz, K. M., Choong, Y. S., Cournia, Z., Isayev, O., Soares, T. A., Wei, G. W., et al. (2024). Machine learning in materials science. J. Chem. Inf. Model. 64 (10), 3959–3960. doi:10.1021/acs.jcim.4c00727

PubMed Abstract | CrossRef Full Text | Google Scholar

Ni, L. K., Ahmad, S. F., Alshammari, T. O., Liang, H. S., Alsanie, G., Irshad, M., et al. (2023). The role of environmental regulation and green human capital towards sustainable development: the mediating role of green innovation and industry upgradation. J. Clean. Prod. 421, 138497. doi:10.1016/j.jclepro.2023.138497

CrossRef Full Text | Google Scholar

Park, J. (2024). Analyzing the direct role of governmental organizations in artificial intelligence innovation. J. Technol. Transf. 49 (2), 437–465. doi:10.1007/s10961-023-10048-4

CrossRef Full Text | Google Scholar

Perez, C. (2002). “Technological revolutions and financial capital: the dynamics of bubbles and golden ages,” in echnological revolutions and financial capital. (Edward Elgar Publishing). doi:10.4337/9781781005323

CrossRef Full Text | Google Scholar

Pintér, L., Kok, M., and Almassy, D. (2017). Measuring progress in achieving the sustainable development goals. GTG: SDGs 1, 99–132. doi:10.7551/mitpress/9780262035620.003.0005

CrossRef Full Text | Google Scholar

Qian, J. L., Zhou, Y. X., and Hao, Q. Y. (2024). The effect and mechanism of digital economy on green total factor productivity—empirical evidence from China. J. Environ. Manag. 372, 123237. doi:10.1016/j.jenvman.2024.123237

PubMed Abstract | CrossRef Full Text | Google Scholar

Rammer, C., Fernández, G. P., and Czarnitzki, D. (2022). Artificial intelligence and industrial innovation: evidence from German firm-level data. Res. Policy 51 (7), 104555. doi:10.1016/j.respol.2022.104555

CrossRef Full Text | Google Scholar

Ran, R., Wang, X. Y., Wang, T., and Hua, L. (2023). The impact of the digital economy on the servitization of industrial structures: the moderating effect of human capital. Data Sci. Manag. 6 (3), 174–182. doi:10.1016/j.dsm.2023.06.003

CrossRef Full Text | Google Scholar

Rasheed, M. Q., Yuhuan, Z., Nazir, M., Ahmed, Z., and Yu, X. (2025). How do semiconductors, artificial intelligence, geopolitical risk, and their moderating effects shape renewable energy production in leading semiconductor manufacturing countries? Technol. Soc. 80, 102761. doi:10.1016/j.techsoc.2024.102761

CrossRef Full Text | Google Scholar

Rasheed, M. Q., and Yuhuan, Z. (2025). Exploring the impact of technology related to artificial intelligence, circular economy, and energy intensity on sustainable development goal 13. Sustain. Dev. doi:10.1002/sd.70274

CrossRef Full Text | Google Scholar

Romer, P. M. (1990). Endogenous technological change. J. Pol. Econ. 22 (5, Part 2), S71–S102. doi:10.1086/261725

CrossRef Full Text | Google Scholar

Singh, A., Kanaujia, A., Singh, V. K., and Vinuesa, R. (2024). Artificial intelligence for sustainable development goals: bibliometric patterns and concept evolution trajectories. Sustain. Dev. 32 (1), 724–754. doi:10.1002/sd.2706

CrossRef Full Text | Google Scholar

Sinha, A., Sengupta, T., and Alvarado, R. (2020). Interplay between technological innovation and environmental quality: formulating the SDG policies for next 11 economies. J. Clean. Prod. 242, 118549. doi:10.1016/j.jclepro.2019.118549

CrossRef Full Text | Google Scholar

Škare, M., Gavurova, B., and Porada-Rochon, M. (2024). Digitalization and carbon footprint: building a path to a sustainable economic growth. Technol. Forecast. Soc. Change 199, 123045. doi:10.1016/j.techfore.2023.123045

CrossRef Full Text | Google Scholar

Spagnuolo, F., Casciello, R., Martino, I., and Meucci, F. (2025). Exploring the impact of artificial intelligence on the pursuit of SDGs: evidence from European state-owned enterprises. Corp. Soc. Responsib. Environ. Manag. 32 (2), 1987–2001. doi:10.1002/csr.3047

CrossRef Full Text | Google Scholar

Truby, J. (2020). Governing artificial intelligence to benefit the UN sustainable development goals. Sustain. Dev. 28 (4), 946–959. doi:10.1002/sd.2048

CrossRef Full Text | Google Scholar

United Nations (2015). Transforming our world: the 2030 agenda for sustainable development. New York, NY: United Nations Resolution. Available online at: https://digitallibrary.un.org/record/3923923?v=pdf.

Google Scholar

United Nations Statistics Division (UNSD) (2019). Global indicator framework for the sustainable development goals and targets of the 2030 agenda for sustainable development. Dev. Sci. Sustain. Dev. Goals Child. Youth 439.

Google Scholar

Vinuesa, R., Azizpour, H., Leite, I., Balaam, M., Dignum, V., Domisch, S., et al. (2020). The role of artificial intelligence in achieving the sustainable development goals. Nat. Commun. 11 (1), 233. doi:10.1038/s41467-019-14108-y

PubMed Abstract | CrossRef Full Text | Google Scholar

Wang, C. L., and Jiao, D. (2025). Impact of artificial intelligence on the labor income distribution: labor substitution or production upgrading? Finance Res. Lett. 73, 106674. doi:10.1016/j.frl.2024.106674

CrossRef Full Text | Google Scholar

Webster, P. (2023). Six ways large language models are changing healthcare. Nat. Med. 29 (12), 2969–2971. doi:10.1038/s41591-023-02700-1

PubMed Abstract | CrossRef Full Text | Google Scholar

Xia, L., Han, Q. J., and Yu, S. (2024). Industrial intelligence and industrial structure change: effect and mechanism. Int. Rev. Econ. Finance 93, 1494–1506. doi:10.1016/j.iref.2024.04.002

CrossRef Full Text | Google Scholar

Xin, B. G., and Ye, X. P. (2024). Robotics applications, inclusive employment and income disparity. Technol. Soc. 78, 102621. doi:10.1016/j.techsoc.2024.102621

CrossRef Full Text | Google Scholar

Xing, Q., Wu, C., Chen, F., Liu, J., Pradhan, P., Bryan, B. A., et al. (2024). Intranational synergies and trade-offs reveal common and differentiated priorities of sustainable development goals in China. Nat. Commun. 15 (1), 2251. doi:10.1038/s41467-024-46491-6

PubMed Abstract | CrossRef Full Text | Google Scholar

Xu, Z. C., Chau, S. N., Chen, X. Z., Zhang, J., Li, Y. J., Dietz, T., et al. (2020). Assessing progress towards sustainable development over space and time. Nature 577 (7788), 74–78. doi:10.1038/s41586-019-1846-3

PubMed Abstract | CrossRef Full Text | Google Scholar

Yang, H. C., Feng, G. F., Gong, Q., and Chang, C. P. (2023). The impact of political competition on green innovation: a new insight into sustainable development. Sustain. Dev. 31 (5), 3692–3708. doi:10.1002/sd.2620

CrossRef Full Text | Google Scholar

Yang, Y., An, R., and Song, J. (2025). Impact of enterprise artificial intelligence on social responsibility: evidence from text analysis. Finance Res. Lett. 75, 106868. doi:10.1016/j.frl.2025.106868

CrossRef Full Text | Google Scholar

Yip, W. (2024). Improving primary healthcare with generative AI. Nat. Med. 30 (10), 2727–2728. doi:10.1038/s41591-024-03257-3

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhang, D. Y. (2024). The pathway to curb greenwashing in sustainable growth: the role of artificial intelligence. Energy Econ. 133, 107562. doi:10.1016/j.eneco.2024.107562

CrossRef Full Text | Google Scholar

Zhang, M. Y., and Yan, J. H. (2021). A data-driven method for optimizing the energy consumption of industrial robots. J. Cleaner Prod. 285, 124862. doi:10.1016/j.jclepro.2020.124862

CrossRef Full Text | Google Scholar

Zhang, X. C., Yu, Q., Liu, J. X., He, Y., and Xu, A. J. (2024). The application of industrial robots and changes in capital-labour income disparity in China. Econ. Analysis Policy 84, 42–56. doi:10.1016/j.eap.2024.08.020

CrossRef Full Text | Google Scholar

Zhao, P., Gao, Y., and Sun, X. (2022). How does artificial intelligence affect green economic growth? Evidence from China. Sci. Total Environ. 834, 155306. doi:10.1016/j.scitotenv.2022.155306

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhao, C. Y., Li, Y. J., Liu, Z. G., and Ma, X. Y. (2024). Artificial intelligence and carbon emissions inequality: evidence from industrial robot application. J. Clean. Prod. 438, 140817. doi:10.1016/j.jclepro.2024.140817

CrossRef Full Text | Google Scholar

Zhou, J. P., Xu, W. X., and Yan, D. (2024a). Breaking the resource curse: the impact of digital economy on the sustainable transformation of resource-based cities. Sustain. Cities Soc. 113, 105707. doi:10.1016/j.scs.2024.105707

CrossRef Full Text | Google Scholar

Zhou, W., Zhuang, Y., and Chen, Y. (2024b). How does artificial intelligence affect pollutant emissions by improving energy efficiency and developing green technology. Energy Econ. 131, 107355. doi:10.1016/j.eneco.2024.107355

CrossRef Full Text | Google Scholar