China’s AI Pilot Zones Show How Smarter Industry Can Raise Output on Limited Land

22.5% improvement in economic output per unit of industrial land following designation as an AI pilot zone, with digital business clusters and public-data policies helping explain the results.

A city can build more factories to grow its economy, but every new industrial site adds pressure to a limited supply of land. China's artificial intelligence pilot zones offer evidence that smarter production and better connections between businesses can help cities generate more economic value from the industrial land they already use, creating another option for places balancing growth with competing demands for space.

Published in Frontiers in Sustainable Food Systems, the research paper "Artificial intelligence policy, digital industry agglomeration, and industrial land intensive use: evidence from a quasi-natural experiment in China" examines this relationship across 271 Chinese cities. Ning Wang, Yunpeng Fu and Zhihong Liu of Liaoning University report an estimated 22.5% improvement in economic output per unit of industrial land following designation as an AI pilot zone, with digital business clusters and public-data policies helping explain the results.

Tracking what changed after cities became AI pilots

The researchers analysed city records from 2010 to 2023, using the phased introduction of national AI innovation and development pilot zones from 2019 as a policy experiment. Their approach compared changes in designated cities with changes in other cities, accounting for persistent local differences, nationwide annual developments and factors such as economic conditions, technology spending, government intervention, finance, wages and industrial structure.

Land-use intensity was measured through secondary-sector value added, covering activities such as manufacturing and construction, divided by industrial land area. This captures the economic value generated per unit of land, giving readers a more precise meaning for "better land use" than a simple count of factories or buildings.

Digital industry clustering was mainly measured through the concentration of employment in telecommunications, computer services, software and related sectors. The main estimate pointed to an average improvement of about 22.5% in pilot cities relative to the comparison trend. Gains generally strengthened after designation; the fourth-year estimate rested on just six early pilot cities with strong digital foundations.

Matching cities with similar characteristics, randomly assigning hypothetical pilot status 1,000 times, changing measures and removing major municipalities broadly preserved the positive finding; allowing for repeated observations within each city reduced the strength of the statistical evidence.

Digital business clusters explain part of the gains

AI policies were associated with a stronger local concentration of digital businesses and workers, and this clustering accounted for approximately 15.53% of the overall estimated effect on industrial land-use intensity. The relationship remained apparent when the researchers used earlier employment data and an alternative measure based on the density of registered information and communications technology companies.

For a manufacturer, proximity to software suppliers, specialist workers and technology firms can make adopting new systems easier and less expensive. The authors suggest that shared knowledge, infrastructure and services help existing businesses improve production, and that the arrival of businesses generating greater value from smaller sites raises an area's economic density. These explanations describe how clustering could work; the city-level analysis does not separately measure every business decision behind the pattern.

Clustering explained only part of the estimated improvement, leaving room for other routes, including production upgrades and changes in the mix of businesses using industrial sites. The authors interpret additional tests as suggesting that output growth was the main driver of the gains, with land-use restructuring also potentially contributing. Their findings do not establish a matching percentage reduction in industrial land area or show how much land any individual company saved.

Open public data helps AI policies deliver more

Cities combining AI pilot policies with open public-data policies showed stronger gains than the model attributed to the two policies separately. Access to government-held data can give businesses more material for developing and applying AI systems, and the analysis found that data openness strengthened both digital industry clustering and its connection with more intensive industrial land use.

Opening a data portal alone did not guarantee better results, with the standalone policy associated with lower land-use intensity in the model. The researchers suggest that compliance costs, incompatible standards, uneven data quality and unreliable interfaces can make newly released information difficult to use. These are proposed explanations for the pattern, underscoring why the practical usability of data matters alongside its availability.

The combined-policy benefit appeared in ordinary prefecture-level cities and higher-tier cities, with a slightly larger estimate for the former. Cities with earlier public-data platforms showed a clearer benefit than those with later platforms or none, suggesting that accumulated datasets and established procedures matter. Tests of the indirect clustering pathway also found clearer evidence when open-data policies accompanied AI support.

Better land policy needs usable data and realistic expectations

The authors recommend extending AI pilots to suitable cities, investing in digital infrastructure such as 5G networks and computing platforms, and helping digital firms develop through skills, financing and business support. Common data standards, dependable sharing systems and coordination between technology and land-planning policies could help businesses put these investments to practical use, with local strategies reflecting each city's starting conditions.

Regular monitoring should track the economic value produced on industrial land and eventually include environmental performance, since a higher output figure cannot describe every consequence of development. The analysis uses an economic-density measure, relies mainly on employment to capture digital clustering and records data openness through platform launch rather than detailed measures of quality or use. Effects on neighbouring cities and individual firms need closer investigation, and the Chinese policy setting limits direct comparisons with other countries.

More productive industrial sites could reduce pressure to expand onto surrounding farmland, giving land-use planners additional options for protecting food production. The paper does not measure farmland preservation, agricultural yields or food-security outcomes, and any such benefit would depend on planning rules and subsequent land decisions. Its strongest practical finding is that AI support, digital business networks and usable public data can work together to help cities generate more economic value from industrial land.