How AI and environmental governance improve air quality in urban skies

A new study has found strong evidence that artificial intelligence (AI) can significantly improve urban air quality by driving both technological innovation and environmental governance. Published in Sustainability (2025), the study titled “Artificial Intelligence and Urban Air Quality: The Role of Government and Public Environmental Attention” uses a quasi-natural experiment design to measure how AI impacts air pollution, identifying critical mediators like government and public environmental attention.

The analysis covers 274 cities from 2011 to 2021 and leverages the rollout of China's Artificial Intelligence Innovation Pilot Zones (AIPZ) as a proxy for AI deployment.

How does AI directly influence urban air quality?

The study has found that the adoption of AI leads to statistically significant improvements in air quality. Using PM2.5 levels as the primary indicator of air pollution, the researchers observed that cities designated as AI pilot zones experienced a reduction in pollution levels. AI contributes directly to pollution control through two main mechanisms: improving energy efficiency and enabling green technological innovation.

AI’s optimization capabilities allow for smarter energy usage across industries, minimizing waste and streamlining processes that traditionally rely on carbon-intensive inputs. By embedding AI into manufacturing and transportation systems, cities can reduce the emissions generated through inefficient operations. Furthermore, the widespread integration of AI supports the development and deployment of cleaner technologies by accelerating research and development cycles. These improvements manifest in lower emissions at the enterprise level and result in better ambient air quality across urban areas.

The study also tested the robustness of this correlation using several methods, including placebo tests and propensity score matching, which confirmed that the observed improvements in air quality were causally linked to AI implementation, not to unrelated urban or policy factors.

What role do government and public environmental attention play?

The impact of AI on air quality does not operate in a vacuum. The study finds that the effectiveness of AI in reducing pollution is significantly enhanced when accompanied by government environmental attention (GEA). In cities where AI technology was adopted, government agencies became more proactive in environmental monitoring, regulatory enforcement, and investment in pollution control. The development of AI increased the frequency of environment-related terms in government work reports, indicating a stronger administrative focus on ecological outcomes.

This heightened governmental focus enables more precise monitoring and better policy formulation. AI helps reduce information asymmetry between administrative departments and facilitates more scientific assessment of regulatory performance. Consequently, local governments invest more in environmental infrastructure, enforce emission standards more rigorously, and guide enterprises toward greener practices, contributing to overall air quality improvement.

In addition to GEA, the study introduces a second moderator: public environmental attention (PEA). Drawing on Baidu search data for haze-related terms, the researchers found that cities with higher levels of public concern about pollution saw even greater benefits from AI deployment. Public awareness and activism amplify governmental responses, exerting bottom-up pressure on officials to prioritize air quality. This dynamic creates a feedback loop where AI-enabled data accessibility enhances transparency, mobilizes citizen engagement, and intensifies environmental governance.

Are these effects uniform across different types of cities?

Not all cities experience AI’s benefits equally. The study highlights significant geographic and structural heterogeneity in outcomes. Eastern cities in China, which are typically more economically advanced, exhibit stronger air quality improvements following AI adoption. These cities already possess the industrial infrastructure, talent base, and funding mechanisms necessary to integrate advanced technologies effectively.

Conversely, mid-western and resource-based cities show comparatively weaker results. In these areas, industrial activity is often dominated by heavy, polluting sectors such as mining and high-energy manufacturing. Limited infrastructure and lower levels of technological readiness impede AI integration, reducing its potential impact on air quality. Additionally, these cities often suffer from path dependency on resource extraction industries, which creates institutional and economic resistance to green transformation. The study suggests that AI alone is insufficient in such environments without complementary reforms and capacity building.

Cities not heavily reliant on resource industries are more agile in deploying AI for environmental improvements. Their diverse economic bases and higher innovation capacity allow AI to drive changes in both industrial processes and public service delivery. These findings indicate that while AI holds transformative potential, its success depends heavily on local conditions, including economic structure, governance capacity, and public engagement.

A blueprint for policy and future research

This study offers clear policy recommendations. First, it calls for regionally tailored strategies to support AI deployment, encouraging cities to build their digital capabilities based on local economic and industrial profiles. Second, it emphasizes the need to embed environmental metrics within official performance evaluations, ensuring that pollution control becomes a governance priority. Third, it advocates for the transfer of advanced technologies from well-resourced eastern cities to underdeveloped regions through state-supported knowledge-sharing mechanisms.

Moreover, the study recommends promoting environmental awareness campaigns to increase public pressure for clean air, which in turn motivates government action. Strengthening public participation is framed not merely as a civic virtue but as a strategic asset in environmental policy.

The authors note limitations in data coverage, particularly the focus on PM2.5 to the exclusion of other pollutants like SO₂ and NO₂. They call for future studies to explore the broader ecological impacts of AI and to refine metrics for evaluating AI adoption across cities. Improved data granularity would allow for more nuanced insights into sector-specific impacts and long-term outcomes.