TY - JOUR

T1 - How Does AI Technology Innovation Drive Carbon Emission Efficiency? A Machine Learning–Based Meta-Frontier Analysis Across 75 Countries

AU - Zhang, F.

AU - Li, R.

AU - Wang, Q.

N1 - Export Date: 29 March 2026; Cited By: 20; Correspondence Address: R. Li; School of Economics and Management, China University of Petroleum (East China), Qingdao, China; email: lirr@upc.edu.cn; Q. Wang; School of Economics and Management, China University of Petroleum (East China), Qingdao, China; email: wangqiang7@upc.edu.cn

PY - 2026/2/1

Y1 - 2026/2/1

N2 - Amid global disparities in technological advancement and carbon emissions, this study evaluates the role of artificial intelligence (AI) technology innovation in improving carbon emission efficiency. AI innovation is assessed using country–year patent data, distinguishing between technology-oriented and application-oriented domains. To measure carbon emission efficiency while accounting for technological heterogeneity across countries, we develop a machine learning–based meta-frontier evaluation framework. This framework provides complementary assessments of efficiency from the perspectives of the meta-frontier, group frontiers, and the technology gap. Results reveal that AI technology innovation significantly improves carbon emission efficiency and narrows technological gaps. Technology-oriented AI exerts stronger effects than application-oriented AI, and the relationship between AI and efficiency follows an inverted U-shape, with the largest gains observed in middle-tier technology groups. Rising income levels further strengthen both the magnitude and persistence of these impacts. Mechanism analysis shows that AI enhances efficiency primarily through technological progress, while regulatory quality and clean energy adoption serve as enabling conditions, and market forces alone remain insufficient. These findings demonstrate that AI can reduce global carbon inequalities, but its sustainability potential depends critically on supportive governance and clean energy transitions. © 2025 ERP Environment and John Wiley & Sons Ltd.

AB - Amid global disparities in technological advancement and carbon emissions, this study evaluates the role of artificial intelligence (AI) technology innovation in improving carbon emission efficiency. AI innovation is assessed using country–year patent data, distinguishing between technology-oriented and application-oriented domains. To measure carbon emission efficiency while accounting for technological heterogeneity across countries, we develop a machine learning–based meta-frontier evaluation framework. This framework provides complementary assessments of efficiency from the perspectives of the meta-frontier, group frontiers, and the technology gap. Results reveal that AI technology innovation significantly improves carbon emission efficiency and narrows technological gaps. Technology-oriented AI exerts stronger effects than application-oriented AI, and the relationship between AI and efficiency follows an inverted U-shape, with the largest gains observed in middle-tier technology groups. Rising income levels further strengthen both the magnitude and persistence of these impacts. Mechanism analysis shows that AI enhances efficiency primarily through technological progress, while regulatory quality and clean energy adoption serve as enabling conditions, and market forces alone remain insufficient. These findings demonstrate that AI can reduce global carbon inequalities, but its sustainability potential depends critically on supportive governance and clean energy transitions. © 2025 ERP Environment and John Wiley & Sons Ltd.

KW - artificial intelligence

KW - carbon emission efficiency

KW - low-carbon development

KW - machine learning

KW - meta-frontier analysis

KW - technological inequality

KW - carbon emission

KW - innovation

KW - sustainability

KW - technological development

UR - https://www.mendeley.com/catalogue/412192b4-82dc-31fb-905c-243d8cc4ec33/

U2 - 10.1002/sd.70312

DO - 10.1002/sd.70312

M3 - статья

VL - 34

SP - 1310

EP - 1349

JO - Sustainable Development

JF - Sustainable Development

SN - 0968-0802

IS - 1

ER -