Measuring the Effects and Examining the Mechanisms of Artificial Intelligence Empowering Rural Revitalization

Abstract

The Chinese government is committed to promoting the deep integration of artificial intelligence (AI) technology with rural development, advocating for the construction of a new “rural revitalization” model supported by information technology. However, its actual effectiveness and economic mechanisms require more thorough empirical verification. Based on Chinese sample data from 2011 to 2023, this study empirically examines the impact of AI on rural revitalization. Findings reveal that AI exerts a significant and sustained supportive role in comprehensive rural revitalization. Deeper analysis reveals that the AI-driven process of comprehensive rural revitalization is not linear but exhibits an inverted “U” pattern—initially promoting then inhibiting development. Mechanism tests indicate that AI empowers comprehensive rural revitalization by enhancing innovation activity and economic dynamism. In the samples where there is a data trading center, a high population density, and a large proportion of rural population, the enabling effect of artificial intelligence is more pronounced, indicating that data elements, population size, and location are important external influencing factors. The paper argues that policy optimization should focus on building a strategic ecosystem for “digital-intelligent villages,” advancing institutional development alongside technological implementation, and devising tiered implementation strategies to fully unleash AI’s enabling potential.

1. Introduction

To eradicate absolute poverty and address increasingly severe challenges such as agricultural marginalization, rural depopulation, and farmer aging, the Chinese government launched the “Rural Revitalization Strategy” in 2017. Its core objective is to drive transformative shifts in production methods [1], enhance governance efficiency, and improve livelihoods across rural areas [2]. However, constrained by location, endowments, and entrenched development patterns, rural areas struggle to achieve endogenous evolution through their own resources alone. External technological impetus and resource infusion are indispensable. Currently, new-generation digital technologies such as artificial intelligence (AI) are rapidly penetrating into all aspects of production and life, providing unprecedented opportunities for achieving comprehensive rural revitalization. They are also becoming a key force to break through the barriers between urban and rural development and activate the internal driving force of rural areas [3,4]. Accordingly, the Chinese government has repeatedly emphasized in its 2025 policy documents the need to “develop smart agriculture and expand application scenarios for technologies like artificial intelligence,” aiming to drive comprehensive rural revitalization and sustainable development through disruptive technological integration.

Existing research typically examines AI’s role in rural revitalization through the lens of technology diffusion. Digital technology has a typical “general-purpose” attribute, which can be widely applied to all aspects of social production. In the past decade, the popularization of digital platforms and mobile internet devices has effectively promoted economic development in rural areas of China. As a representative of the new generation of digital technologies, artificial intelligence does not merely achieve basic digital access; instead, it leverages unique technical advantages such as precise supply–demand matching, intelligent decision-making prediction, and dynamic resource scheduling to upgrade the efficiency of existing digital technologies [5,6], further resolving the deep-seated problems of blocked economic element circulation and inefficient resource allocation in rural areas, and injecting more targeted and empowering technological impetus into rural economic development. Its transfer from information and communication fields to agricultural production significantly enhances social productivity, empowering all aspects of rural production and giving rise to new forms of modern agriculture. While broadening the scope of agricultural labor and increasing rural incomes [3], it also creates more employment opportunities and fosters a more livable environment in rural areas [7]. Some scholars also believe that digital technologies such as artificial intelligence have extremely strong penetrability and renewability, capable of integrating with traditional agricultural technology frameworks to expand agriculture’s economic and social functions. This fundamentally alters the logic of traditional agriculture, driving the integrated development of agriculture with secondary and tertiary industries [8,9,10]. In summary, existing research has generally supported the positive role of digital technologies such as artificial intelligence in the comprehensive revitalization of rural areas, broadly supports the positive role of AI in comprehensive rural revitalization, indicating that AI development drives all-round revitalization of rural industries, culture, ecology, organizations, and talent [11].

This study has a strong basis because previous research has examined the relationship between AI and comprehensive rural rehabilitation from a variety of angles. New data is still required, though, in areas like evaluating nonlinear correlations, creating quantitative measurement frameworks, and expanding our knowledge of underlying systems. In particular, first, in a general sense, the role of digital technology in rural revitalization was analyzed, but there was a lack of discussion on the characteristics of artificial intelligence technology and the deep logical relationship between it and the comprehensive revitalization of rural areas. As a result, the clarification of the transmission mechanism was rather broad. Second, existing research mainly focuses on the opportunities brought by the application of artificial intelligence technology for the development of rural areas in China. Existing research has not provided an answer. Third, the existing research mainly adopts qualitative analysis methods and has not yet constructed an effective statistical measurement index system, nor has it used quantitative analysis tools to support relevant arguments, which limits the further discussion and expansion of relevant topics. Third, the existing studies mostly use static analysis framework for the discussion of artificial intelligence and comprehensive rural revitalization and lack the induction and summary of nonlinear relationships such as dynamic evolution characteristics and inflection point effect, resulting in relatively single conclusions.

This paper makes contributions in several key areas: First, on the basis of the existing research on digital technology and rural revitalization, this paper focuses on the deep impact of artificial intelligence technology on the comprehensive rural revitalization, summarizes and combs the logical relationship between the two from the perspective of the unique technical attributes of artificial intelligence and the main difficulties faced by rural revitalization, and gives explanations based on China’s economic reality. Second, it posits that enhancing innovation dynamism and economic vitality constitutes a crucial internal mechanism for AI’s enabling effects, while population density and data elements serve as key external influencing factors. This establishes a more comprehensive theoretical analytical framework from both internal and external dimensions. Third, based on the connotation of artificial intelligence and comprehensive rural revitalization, a more comprehensive statistical measurement index system was constructed. Using tools such as dynamic effect models, quadratic term models, and quantile models, the causal relationship and nonlinear characteristics between artificial intelligence and comprehensive rural revitalization were deeply examined. This answered the boundary and adaptability issues of artificial intelligence technology investment, providing more sufficient empirical evidence for the in-depth exploration of related research. Fourth, integrating theoretical and empirical findings, this paper proposes actionable policy recommendations from institutional design and implementation perspectives, aiming to offer valuable insights for developing countries addressing rural development challenges.

The rest of the paper is arranged as follows: Section 2 is the mechanism analysis and research hypothesis, which mainly analyzes the process of artificial intelligence enabling rural revitalization from the two levels of mechanism and transmission path; Section 3 is the research design, including index setting, variable selection, data source and other contents; Section 4 is the empirical analysis, which mainly presents the empirical test results of causal analysis, mechanism analysis, nonlinear relationship test and so on; and Section 5 is the conclusion and policy enlightenment, which mainly reports the basic conclusions of this paper and the policy enlightenment contained in it.

2. Mechanism Analysis and Research Hypotheses

The core economic mechanism of AI empowering the comprehensive revitalization of the countryside lies in systematically solving the problem of unbalanced and inadequate rural development by reshaping the quality and combination of rural production factors and promoting the comprehensive upgrading of agriculture, the comprehensive progress of rural areas, and the comprehensive development of farmers. The traditional rural economic growth model is restricted by the law of diminishing marginal returns of primary factors and the linear evolution trajectory of simple technologies, and it is easy to fall into a vicious circle of high investment, low efficiency, and unsustainable development. This is an important reason for the long-term backward [12] development of rural areas in China [13]. It is also for this reason that the relatively closed rural economic system cannot rely on its own resource endowment to achieve endogenous iteration and must rely on external technology promotion and resource input.

So why can artificial intelligence drive comprehensive revitalization and sustainable development in rural economies? Scholars like [14] argue that digital technologies such as AI possess a quintessential “general-purpose” nature. Typically centered on data-driven approaches, algorithmic iteration [15] and model optimization, these technologies can be widely applied across all aspects of rural social production [6]. On one hand, AI revolutionizes agricultural tools through intelligent machinery and digital platforms, reshaping the nature of rural production factors. This promotes the intensification, efficiency, and precision upgrades of rural industries, mitigates constraints imposed by natural conditions and technological barriers on rural production activities, enhances farmers’ digital skills and smart equipment operation capabilities, and drives the adoption of new agricultural technologies. On the other hand, AI technology expands the boundaries of modern agriculture by transforming agricultural data into production factors. This broadens agricultural production frontiers and industrial chains, accelerates the integrated development of agriculture with secondary and tertiary industries, reconfigures the combination patterns and value of rural production factors, and solidifies the material foundation for rural industrial revitalization. Furthermore, unlike traditional digital platforms which only perform information transmission and basic transaction functions in rural areas, artificial intelligence has more obvious technical advantages in data mining, precise matching, and dynamic prediction. This helps rural areas fully leverage the value of digital infrastructure, achieving functional upgrades and value increments based on traditional digital platforms, breaking through barriers in the flow of data elements between urban and rural areas, and promoting the intelligent upgrade of rural public services and the precision of ecological governance [16]. This provides comprehensive support across dimensions such as industry [17], talent, ecology [3], and governance for achieving the goal of comprehensive rural revitalization [4].

The enabling role of AI in comprehensive rural revitalization is not static but exhibits dynamic evolution and continuous deepening over time. As a production system characterized by high complexity and path dependency, the rural economy faces both adaptation challenges and incremental breakthroughs in absorbing exogenous technologies—a phenomenon particularly evident in AI’s penetration. Initially constrained by the inertia and path dependency of existing growth models, AI’s empowerment primarily addresses superficial issues in agricultural production. Consequently, initial resistance and adaptation issues arise, limiting its impact on comprehensive rural revitalization. As technology continues to permeate and embed itself, AI drives profound restructuring of production factors. Innovative allocation occurs between data elements and traditional factors like land and capital, while widespread adoption of agricultural big data platforms and intelligent decision-making systems propels agricultural production from “experience-driven” to “data-driven” [18]. This endogenous transformation further spurs the emergence of new service models like agricultural product traceability. It deepens integration with rural industries, ecosystems, and governance, giving rise to cross-sectoral innovations such as agritourism and biomanufacturing. This drives efficiency gains and value expansion across rural industrial chains [19]. Throughout this process, artificial intelligence technologies form self-reinforcing mechanisms through data-driven evolution, algorithmic refinement, and model iteration. Their computational power and analytical depth grow exponentially, establishing a positive feedback loop with the rural economic system characterized by “adaptation–integration–extension.” This cycle continuously empowers comprehensive rural revitalization.

In summary, this study proposes the following hypothesis:

H1: 

Artificial intelligence exerts a significant, dynamic, and sustained promotional effect on comprehensive rural revitalization.

While existing research confirms AI’s multi-dimensional empowerment of rural revitalization, some scholars argue that excessive AI investment may trigger negative consequences—the productivity paradox [20]—due to technological application lags and institutional development delays. This economic principle remains valid within the context of comprehensive rural revitalization. Within a reasonable input threshold, the moderate application of artificial intelligence can release a significant positive enabling effect, which can not only effectively replace repetitive and high-intensity labor in agricultural production and liberate rural productivity, but also meet the multi-dimensional needs of rural industrial development and grassroots governance based on its technical characteristics, becoming an important driving force to promote the comprehensive revitalization of rural areas. This is also the core reason for the continuous rise in the enabling effect in the inverted U-shaped relationship. However, once the input threshold is exceeded, excessive technology input will shift from promotion to inhibition. On the one hand, if the scale of labor substitution is blindly increased without independent external evaluation, there will be decision-making errors in the scale of labor substitution, leading to serious undesirable labor substitution, resulting in a large number of idle rural labor, and the employment channels in rural areas are already narrow. At the same time, if there is a lack of auxiliary measures such as skills training for idle labor, the cultivation of emerging rural industries, and the docking of employment positions, idle labor cannot be timely transferred to new employment fields, which will not only cause a serious waste of human resources, but also impact the traditional rural social structure and employment ecology, and inhibit the growth of farmers’ income and the improvement of people’s livelihood. On the other hand, the rural economy covers agricultural production, rural commerce, social governance and other fields. Agricultural production has the characteristics of natural dependence and geographical dispersion, while farmers’ operation is small and scattered. The demand for technology is characterized by fragmentation and diversification, which is in natural contradiction with the systematic and integrated application requirements of artificial intelligence technology itself. Under excessive investment, complex intelligent systems are difficult to accurately adapt to the actual scene of decentralized and fragmented rural areas, which is easy to produce problems such as high cost of technology promotion and low efficiency of implementation, and even form “technology suspension” divorced from the actual rural areas, which cannot play the due effect of technology empowerment, and will crowd out other high-quality resources for rural development due to resource misallocation. It further hinders the comprehensive revitalization process of rural industrial upgrading, ecological optimization and governance improvement, and finally makes the enabling effect of artificial intelligence fall into a dilemma from positive to negative, thus forming an inverted “U” shaped relationship characteristic of promotion first and then inhibition. In summary, this study proposes the following hypothesis:

H2: 

A significant inverted U-shaped relationship exists between artificial intelligence and comprehensive rural revitalization.

Traditional rural economic growth models are constrained by the linear evolution trajectory lock-in of simple technologies [10]. Production processes suffer from fragmentation and knowledge lag, resulting in severe deficiencies in innovation drivers and iterative capabilities. This limits further improvements in agricultural production efficiency and value-added potential. Leveraging algorithmic iteration and data processing advantages, AI can deeply mine data across the entire agricultural production process. This enables the identification of innovation clues hidden within massive information, providing precise insights into optimizing agricultural production and management models, as well as pathways for adding value to agricultural products. AI transforms fragmented experience into systematic innovative knowledge, activating sources of technological innovation. Moreover, AI lowers the barriers to using digital platforms (such as smart supply chains and e-commerce live-streaming systems), breaks down information barriers between urban and rural markets, narrows the “digital divide,” and creates scenarios for rural entities to interact with diverse stakeholders. This sparks business model innovation through demand matching and resource integration, giving rise to new models like smart agriculture and rural cultural-tourism integration. It drives the coordinated advancement of talent, culture, ecology, and organizational revitalization, ultimately achieving comprehensive rural revitalization. In summary, this paper proposes the following research hypothesis:

H3: 

Artificial intelligence empowers comprehensive rural revitalization by enhancing innovation activity.

The rural economy has long been constrained by the obstruction of factor mobility and the singularity of industrial forms, resulting in a much lower economic circulation activity than that in urban areas. Although digital platforms have initially achieved the connection between rural resources and the market, traditional digital platforms only fulfill the functions of information transmission and basic transactions in rural areas, and have inherent limitations such as insufficient resource matching accuracy, lagging market response, and low efficiency in factor allocation. They are unable to fundamentally solve the static circulation problem of the rural economy. However, artificial intelligence, with its unique technical advantages in data mining, precise matching, and dynamic prediction, has achieved functional upgrades and value addition on the basis of traditional digital platforms, providing indispensable technical support for enhancing the activity of the rural economy. Its additional value mainly lies in two aspects: Firstly, the rural e-commerce live-streaming system and smart supply chain driven by artificial intelligence do not simply realize “online transactions,” but precisely match supply and demand through user profiles, predict market demand based on consumption data, and optimize logistics scheduling and inventory management through intelligent algorithms. This not only breaks the physical spatial limitations of the economy but also achieves efficient and precise connection of rural characteristic products, cultural tourism resources with the national and even global markets, promoting the transformation of agricultural products from “passive waiting for orders” to “precise order-driven production,” significantly improving the market response speed and adaptability of rural industries. Secondly, the new types of digital tourism and other new business models enabled by artificial intelligence, compared to the traditional online promotion model of tourism, can precisely reach potential customers through intelligent recommendation algorithms, enhance consumption experience through scene simulation technology, and dynamically optimize business layout through big data analysis. They attract urban capital and consumption power to flow to the countryside more efficiently, precisely match rural job demand and talent skills characteristics, promote the directional concentration and efficient reorganization of funds, technologies, and talents, and strengthen the dynamic adaptability of factor allocation. This technology-enabled empowerment model based on the additional value of AI breaks the functional boundaries of traditional digital platforms, promotes the transformation of the rural economy from “static circulation” to “dynamic value-added,” and thus consolidates the foundation of the rural economy, drives industrial revitalization, talent return, and governance optimization, and injects sustainable economic momentum into the comprehensive revitalization of the countryside. In summary, this paper proposes the following research hypothesis:

H4: 

Artificial intelligence empowers comprehensive rural revitalization by enhancing economic activity.

3. Research Design

3.1. Variable Description

3.1.1. Dependent Variable: Comprehensive Rural Revitalization (Rural)

According to documents issued by the Chinese government, including the “Opinions of the Central Committee of the Communist Party of China and the State Council on Implementing the Rural Revitalization Strategy” (https://www.gov.cn/zhengce/2018-02/04/content_5263807.htm (accessed on 17 January 2026)) and the “Comprehensive Rural Revitalization Plan (2024–2027)” (https://www.gov.cn/gongbao/2025/issue_11846/202502/content_7002798.html (accessed on 17 January 2026)), comprehensive rural revitalization is described as the all-round revitalization of rural areas encompassing industries, ecology, talent, culture, and organizations. Its ultimate goals are thriving industries, a pleasant ecological environment, a civilized rural ethos, effective governance, and affluent living standards. Centered on this conceptual framework, the measurement framework for “Rural Comprehensive Revitalization” (Rural) in this paper comprises five primary indicators—including thriving industries—and 21 secondary indicators—such as total agricultural machinery power. These are synthesized using the “entropy weighting method.” Detailed indicator composition is presented in Table 1.

Table 1. Measurement Index System for the Revitalization of All Villages in China.

Primary Indicators	Secondary Indicators	Unit/Calculation Method
Level of Agricultural Development	Total Agricultural Machinery Power (+)	10,000 kW
	Land Productivity (+)	Total Output Value of Agriculture, Forestry, Animal Husbandry, and Fisheries/Effectively Irrigated Area
	Comprehensive Grain Production Capacity (+)	Grain Yield per Unit Area (kg/ha)
	Labor Productivity (+)	Value Added of Primary Industry/Rural Population
	Rural electricity consumption (+)	Billion kilowatt-hours
	Pesticide usage (−)	10,000 tons
Farmers’ standard of living	Rural residents’ income level (+)	yuan
	Rural residents’ consumption level (+)	yuan
	Average Number of Passenger Vehicles Owned per 100 Rural Households at Year-End (+)	vehicles
	Per capita residential floor area for rural residents (+)	square meters
Rural public service capacity level	Rural water supply coverage rate (+)	%
	Environmental health service level (+)	Number of Village Health Clinics (units)
	Township cultural stations (+)	units
	Average Years of Education for Rural Population (+)	Total converted education years/Population aged 6 and above
	Per capita expenditure on culture, education, and recreation among rural residents (+)	yuan
	Rural television program population coverage rate (+)	%
Urban-rural development balance	Urbanization rate (+)	Urban Population/Year-End Permanent Residents
	Growth Rate of Per Capita Income for Farmers (+)	Income Growth/Previous Year’s Income
	Rural Household Engel’s Coefficient (−)	%
	Urban-rural income gap (−)	Urban-rural per capita disposable income gap
	Urban-rural consumption gap (−)	Urban-rural per capita consumption expenditure gap

Note: “+” and “−” in parentheses denote positive and negative indicators in the entropy weighting method.

3.1.2. Explanatory Variable: Artificial Intelligence (AI)

Following the methodology of [21], the explanatory variable “Artificial Intelligence” in this study is measured using the number of relevant enterprises in each province (municipality) during the sample year. According to the definition of existing research, “AI enterprises” refer to enterprises that are committed to the research and development and promotion of artificial intelligence-related technologies such as machine learning and image recognition. According to the definition of the concept of “artificial intelligence” in the above research, the keywords of the business scope of enterprises are reviewed and analyzed, and all enterprises involved in machine learning, deep learning, chips, image recognition and other artificial intelligence-related technologies are classified, counted and linearly interpolated according to the place of ownership, year and other information, and named as “AI” after taking the logarithm.

3.1.3. Mediating Variables

The first mediator variable is “Innovation Activity”, This index refers to the cumulative value of the agreed technology transaction price (including the actual valuation amount of various technology transactions such as technology development, transfer, consultation and service) in various technology contracts during the reporting period. It is considered as a “barometer” of regional innovation activity. After taking the logarithm, it is named “Innovation_ac”.

The second mediating variable is “Economic Activity”, measured by the number of legal entities engaged in primary industries within the sample year for each province (municipality). This indicator refers to organizations legally established and independently engaged in agricultural production (forestry, animal husbandry, and fisheries), typically used to characterize the scale and activity level of a region’s primary industries. After taking the logarithm, it is named “Economic_ac”.

3.1.4. Control Variables

Considering that the explanatory variable “agricultural revitalization” in this article is a comprehensive indicator system, and it is affected by other potential factors in addition to artificial intelligence, in order to reduce the estimation bias caused by the omission of variables, the selection of control variables in this article fully takes into account other contents related to agricultural revitalization, mainly including factors such as economic development level, population size, fiscal expenditure, and informatization level. Specifically: Science expenditure (Science) is measured by the annual total science expenditure of each province (municipality directly under the central government), referring to government funding allocated for developing and promoting various scientific research endeavors, covering both natural and social sciences. Education expenditure (Education) is measured by the annual total education expenditure of each province (municipality directly under the central government), referring to all government spending on education, including capital investment in educational infrastructure and recurrent educational expenses. Capital is measured using the year-end balance of loans from financial institutions in each province (municipality). Information level is measured using the year-end number of mobile phone users in each province (municipality). Missing values were imputed using linear interpolation and log-transformed. For detailed information of the above variables, please refer to Table 2.

Table 2. Descriptive Statistics.

Variable Name	English Abbreviation	Observed Value	Mean	Standard Error	Minimum	Maximum
Rural Revitalization	Rural	403	−1.2596	0.3192	−2.1586	−0.6291
Artificial Intelligence	AI	403	8.0301	1.6232	3.2189	12.1846
Regional Gross Domestic Product	GDP	403	9.8115	1.0113	6.4159	11.8180
Population Size	Pep	403	8.1324	0.8415	5.7333	9.4498
Science Expenditures	Science	403	12.2377	1.4904	7.9427	15.4795
Education Expenditures	Education	403	14.1480	0.9104	12.0923	16.3234
Capital Stock	Capital	403	18.6792	0.9172	14.9663	20.8058
Level of Informatization	Information	403	6.9172	0.7934	4.4067	8.4325
Innovation Activity	Innovation_ac	403	10.7495	1.5555	5.1358	13.8090
Economic Activity	Economic_ac	403	4.8805	2.2515	−3.2189	9.0522

Note: All variables above have undergone logarithmic transformation.

3.2. Data Sources

Data for the above variables are sourced from authoritative statistical yearbooks and official databases, primarily from two categories: First, indicators related to agriculture and rural areas are mainly sourced from databases such as the China Statistical Yearbook, China Rural Statistical Yearbook, and statistical bulletins of provinces (municipalities directly under the central government), used to calculate indicators related to rural revitalization; Second, provincial-level macroeconomic indicators, primarily sourced from the China Population and Employment Statistical Yearbook, China Financial Yearbook, the National Enterprise Credit Information Publicity System, and the National Bureau of Statistics website, China Economic and Financial Research Database (CSMAR). Considering data availability and balance, the time window is set from 2011 to 2023. The sample covers China’s 31 provincial-level administrative units (including municipalities directly under the central government; data for Hong Kong, Macao, and Taiwan are currently unavailable). Missing data are imputed using linear interpolation.

3.3. Model Construction

3.3.1. Baseline Model

To minimize interference from individual variability and temporal trends on regression results, this study employs a dual fixed-effects model for both time and individuals. The specific model form is as follows:

$${\mathit{Rural}}_{\mathrm{it}} ={ \mathrm{\alpha }}_0 + {\mathrm{\alpha }}_1{\mathit{AI}}_{\mathrm{it}} +{ \mathrm{\alpha }}_2{\mathit{X}}_{\mathrm{it}} + {\mathrm{\mu }}_{\mathrm{i}} +{ \mathrm{\gamma }}_{\mathrm{t}} + {\mathrm{\epsilon }}_{\mathrm{it}}$$

(1)

where variable subscript $t$ denotes year (time), $i$ denotes province (individual), ${\mathrm{\alpha }}_0$ denotes the intercept term, ${\mathrm{\alpha }}_1$ denotes the coefficient estimate for the explanatory variable, $X_{\mathit{it}}$ denotes the set of control variables, ${\mu }_i$ denotes the individual effect, ${\gamma }_t$ denotes the time effect, and ${\epsilon }_{\mathit{it}}$ denotes the random error term.

3.3.2. Dynamic Effects Model

To estimate the dynamic effects of AI-enabled comprehensive rural revitalization, this study employs the following model, where $\mathit{Dum} \mathit{year}$ denotes the year dummy variable, and the remaining variables are identical to those in Model (1).

$${\mathit{Rural}}_{\mathrm{it}}={ \mathrm{\beta }}_0+{\mathrm{\beta }}_1{\sum }_{\mathrm{t}=2011}^{2023}{(\mathit{AI}}_{\mathrm{it}}·\mathit{dumyear})+{ \mathrm{\beta }}_2{X}_{\mathrm{it}}+{ \mathrm{\mu }}_{\mathrm{i}}+{\mathrm{\epsilon }}_{\mathrm{it}}$$

(2)

3.3.3. Quadratic Regression Model

To examine the inverted U-shaped nonlinear relationship between artificial intelligence and rural revitalization, this paper employs the following quadratic regression model for estimation.

$${\mathit{Rural}}_{\mathrm{it}}={\mathrm{\eta }}_0+{ \mathrm{\eta }}_1{\mathit{AI}}_{\mathrm{it}}+{ \mathrm{\eta }}_2 {\mathit{AI}}_{\mathit{it}}^2+{\mathrm{\eta }}_3 X_{\mathrm{it}}+{\mathrm{\mu }}_{\mathrm{i}}+{\mathrm{\gamma }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{it}}$$

(3)

3.3.4. Mechanism Testing Model

This paper employs the following model for mechanism analysis, where $M_{\mathrm{it}}$ represents the mediating variables (specifically “innovation activity” and “economic activity” in this study), while the remaining variables are identical to those in Model (1).

$$M_{\mathrm{it}}={ \mathrm{\omega }}_0+{\mathrm{\omega }}_1{\mathit{AI}}_{\mathrm{it}}+{\mathrm{\omega }}_2{X}_{\mathrm{it}}+{ \mathrm{\mu }}_{\mathrm{i}}+{\mathrm{\gamma }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{it}}$$

(4)

4. Empirical Analysis

4.1. Benchmark Regression

By progressively incorporating explanatory variables, control variables, time effects, and individual effects, the full results of the benchmark regression are presented in Table 3. Column (1) shows that the coefficient estimate for the explanatory variable is approximately 0.1556, significant at the 1% level. After further incorporating control variables and controlling for individual and time effects (Column 3), the coefficient remains significantly positive, indicating that artificial intelligence exerts a significant positive effect on rural revitalization. Economically speaking, this finding indicates that the development of artificial intelligence and the diffusion of related technologies have reshaped the nature and combination patterns of rural production factors. This promotes the upgrading of rural industries toward intensification, efficiency, and precision, enhancing rural residents’ income levels and living environments while also bringing more development opportunities to rural areas. Consequently, it comprehensively empowers the full revitalization of rural areas across dimensions such as industry, talent, ecology, and governance, helping rural regions achieve thriving industries, ecological livability, cultural vitality, effective governance, and affluent living.

Table 3. Benchmark Regression Results.

Variable	(1)	(2)	(3)
	Rural	Rural	Rural
AI	0.1556 ***	0.0477 ***	0.0578 **
	(0.0061)	(0.0131)	(0.0245)
GDP		0.2753 ***	−0.0131
		(0.0753)	(0.0881)
Pep		−0.1859	0.2865
		(0.2098)	(0.2052)
Science		0.0061	−0.0047
		(0.0132)	(0.0133)
Education		−0.0434	−0.1040 ***
		(0.0368)	(0.0393)
Capital		0.0674 *	0.0245
		(0.0385)	(0.0376)
Information		0.0934 **	0.0256
		(0.0391)	(0.0417)
Constant	−2.5093 ***	−4.6347 ***	−3.1120 **
	(0.0523)	(1.4476)	(1.3726)
Time Effect	No	No	Yes
Individual Effect	No	No	Yes
Observed value	403	403	403
R-squared	0.6263	0.9403	0.9467

Note: For variable definitions, please refer to Table 2, values in parentheses are standard errors; *, **, *** correspond to 10%, 5%, and 1% significance levels, respectively. The remainder of the table follows the same convention.

4.2. Endogeneity Issues and Robustness Tests

4.2.1. Discussion on Endogeneity

Since the explanatory variable used in this article is the number of artificial intelligence enterprises, from the perspective of variable composition, it is prone to be associated with other potential factors, leading to an endogeneity problem in the model. To address this issue, this section employs Generalized Difference-in-Moments Estimation (referred to as Difference-GMM) for processing. The fundamental principle of this method is to utilize lagged terms of the explanatory variables as instrumental variables for corresponding variables in the difference equation, thereby addressing endogeneity issues and obtaining consistent estimators. As shown in the regression results of Table 4, after applying the Difference-GMM estimation, the sign of the coefficient estimates for the explanatory variables remained unchanged, and their significance levels were still maintained at a high level.

Table 4. Robustness Test: Generalized Difference-in-Moment Estimation.

Variable	Rural
AI	0.0182 ***
	(0.0059)
L.Rural	0.8465 ***
	(0.0395)
Constant	−0.2984 ***
	(0.0973)
Control Variables	No
Observations	341
Pseudo R-squared	0.9804

Note: For variable definitions, please refer to Table 2. “L.” denotes the one-period lag. Since this estimation method does not report the R-squared, the Pseudo R-squared is calculated instead, *** correspond to 1% significance level.

4.2.2. Robustness Test: Variable Coefficient Model

Since this study employs provincial panel data as the sample, the fixed-effects model cannot fully eliminate endogeneity issues caused by individual differences. Therefore, it is necessary to further control for the impact of individual fixed effects. To this end, this section conducts regression analysis using a random-effects model. This method assumes that each individual (province) has an independent slope, subsequently controlling for individual effect biases by generating interaction terms between the explanatory and response variables. Table 5 regression results indicate that AI’s enabling effect on comprehensive rural revitalization remains significant, confirming the robustness and reliability of the benchmark regression findings.

Table 5. Regression results of the variable coefficient model.

Variable	Rural
AI	0.0893 ***
	(0.0305)
Constant	−5.0490 ***
	(0.5852)
Control Variables	Yes
Time effect	Yes
Individual effects × Explanatory variables	Yes
Observed values	403
R-squared	0.9308

Note: For variable definitions, please refer to Table 2. *** correspond to 1% significance level.

4.2.3. Robustness Test: Data Truncation

To mitigate the impact of outliers on regression results, the sample data underwent two-tailed 1% trimming. Model (1) was then reapplied for regression analysis. Table 6 shows that after trimming, the significant impact of AI on comprehensive rural revitalization persists, further validating the reliability of the baseline regression findings.

Table 6. Data Truncation.

Variable	(1)	(2)	(3)
	Rural	Rural	Rural
AI	0.1568 ***	0.0473 ***	0.0603 ***
	(0.0063)	(0.0131)	(0.0223)
Constant	−2.5190 ***	−6.3102 ***	−3.7197 **
	(0.0535)	(1.4662)	(1.5124)
Control Variables	No	Yes	Yes
Time effect	No	No	Yes
Individual effect	No	No	Yes
Observed value	403	403	403
R-squared	0.6238	0.9407	0.9476

Note: For variable definitions, please refer to Table 2, **, *** correspond to 5%, and 1% significance levels.

4.3. Dynamic Effect Analysis

Table 7 presents the results of the dynamic effect analysis based on Model (2). Overall, the coefficient estimates for the interaction term (AI × Year) are positive across the entire sample period. This positive correlation is replicated across different time dimensions, supporting the statistical conclusion that a dynamic relationship exists between artificial intelligence and comprehensive rural revitalization. Phased analysis reveals a dynamic and progressively intensifying process of AI empowering rural revitalization. During the initial phase (2011–2017), the coefficient remained positive within a relatively stable range, reflecting the early stages of AI integration into rural socioeconomic systems. This period involved “technology implantation and scenario adaptation,” where AI began enhancing agricultural productivity and facilitating basic information flow, injecting technological momentum into rural development. After 2018, with the implementation of policies such as the National Rural Revitalization Strategy Plan (2018–2022) (https://www.gov.cn/zhengce/2018-09/26/content_5325534.htm (accessed on 17 January 2026)), the synergy between AI and the rural revitalization strategy increased. Technological applications expanded from the production end to the entire industrial chain, forming a “technology + institutional” synergy with policy support. This amplified the role of AI in revitalizing rural industries and optimizing factor allocation, driving further growth in the coefficient. In the later period (post-2020), the coefficient fluctuated but remained significantly positive, reflecting AI’s entry into a “deep integration–efficiency release” phase in rural development. AI continues to reshape rural production and living patterns, empowering comprehensive rural revitalization across economic, social, and ecological dimensions. This confirms that AI’s positive impact on rural revitalization is not static but dynamically reinforced through technological iteration, policy advancement, and systemic adaptation. This trajectory aligns with the evolution of rural revitalization from “strategic planning” to “deepened implementation,” thus validating H1.

Table 7. Dynamic Effects.

Variable	Rural	Variable	Rural
AI × Year2011	0.1877 ***	AI × Year2018	0.1721 ***
	(0.0222)		(0.0205)
AI × Year2012	0.1859 ***	AI × Year2019	0.1706 ***
	(0.0219)		(0.0203)
AI × Year2013	0.1859 ***	AI × Year2020	0.1605 ***
	(0.0219)		(0.0196)
AI × Year2014	0.1849 ***	AI × Year2021	0.1554 ***
	(0.0214)		(0.0189)
AI × Year2015	0.1824 ***	AI × Year2022	0.1490 ***
	(0.0209)		(0.0183)
AI × Year2016	0.1786 ***	AI × Year2023	0.1501 ***
	(0.0206)		(0.0180)
AI × Year2017	0.1754 ***	Constant	−4.4367 ***
	(0.0205)		(1.3606)
Control Variables	Yes	Observation	403
Individual Effects	Yes	R-squared	0.9491

Note: For variable definitions, please refer to Table 2. *** correspond to 1% significance level.

4.4. Testing the Inverted U-Shaped Relationship

To examine the inverted U-shaped relationship between artificial intelligence and comprehensive rural revitalization, this study adopts the methodology proposed by Lind and Mehlum [22]. This approach comprises three key steps: ① establish a quadratic model and confirm the significant negative coefficient of the quadratic term to determine the fundamental curve shape; ② verify opposite slope signs on either side of the inflection point (positive left, negative right) to clarify dynamic characteristics; ③ finally, calculate the inflection point confidence interval using the FIELLER method to ensure interpretable economic significance. Table 8 reports the results of the inverted U-shaped relationship test. The findings indicate that after regression using Model (3), the coefficient estimate for the squared term of the explanatory variable (AI²) is significantly negative, preliminarily indicating that the relationship between artificial intelligence and comprehensive rural revitalization is non-linear, with a pronounced inflection point effect. Further calculations reveal the “inflection point” value to be 8.9547. By measuring the slopes on both sides of this inflection point, the average slope on the left side is found to be 0.0986, while the average slope on the right side is −0.0270. Based on this, the inflection point can be identified as the vertex of an inverted U-shaped curve, falling within the 95% confidence interval (lower bound 7.7709, upper limit 10.2876).

Table 8. U-shaped Relationship Test.

Variable	Rural
AI	0.3466 ***
	(0.0356)
AI2	−0.0194 ***
	(0.0018)
Inverted U-shaped peak value	8.9547
Slope of AI on the left side of the vertex	0.0986
Slope of AI on the right side of the vertex	−0.0270
95% confidence interval for the FLELLER method	[7.7709, 10.2876]
Constant	−5.4330 ***
	(1.1683)
Control Variables	Yes
Time effect	Yes
Observations	403
R-squared	0.9639

Note: For variable definitions, please refer to Table 2. *** correspond to 1% significance level.

The above analysis results show that when the input of artificial intelligence technology is in a reasonable range, the marginal enabling effect of the input is positive. For every unit increase in the development level of artificial intelligence, the level of comprehensive rural revitalization will increase by 0.0986 units on average, indicating that moderate technology input can effectively replace repetitive and high-intensity labor in agricultural production and liberate rural productivity. In addition, the technical system, rural decentralized production and operation, and grassroots governance scenarios have not broken through the boundary of adaptation, and the positive enabling effect has been fully released. However, when the AI development level breaks through the index-point value, the marginal effect of technology investment turns from positive to negative. For each additional unit of AI development level, the level of comprehensive rural revitalization decreases by 0.0270 units on average, which means that excessive technology investment has broken through the adaptation boundary of the rural economic and social system. As a result, problems such as unwanted labor substitution, imbalance of adaptation between technology and rural scene, resource misallocation, and the lack of independent external inspection of artificial intelligence performance, labor transformation and supporting measures for technology localization further amplify the productivity paradox such that the energy efficiency of artificial intelligence should shift from promotion to inhibition. In summary, H2—“A significant inverted U-shaped relationship exists between AI and comprehensive rural revitalization”—is established.

4.5. Mechanism Analysis

4.5.1. Innovation Activity

Traditional rural production processes suffer from constraints such as fragmentation and knowledge lag, resulting in severe deficiencies in innovation momentum and iterative capabilities. These limitations hinder further improvements in agricultural production efficiency [14] and value-added output. Artificial intelligence (AI) represents a quintessential “general-purpose” technology [23]. It can be widely applied across all segments of social production, lowering information access barriers in rural areas. By facilitating demand matching and resource integration, AI sparks business model innovation, giving rise to new paradigms such as smart agriculture and the integration of rural cultural tourism. Simultaneously, the widespread adoption of AI technology significantly lowers the barriers and costs of technological iteration, narrowing the “technology access gap” and “technology iteration gap” between urban and rural areas. This drives the transformation and modernization of rural governance models [16]. Therefore, AI serves as a catalyst for enhancing innovation activity, thereby synergistically advancing talent, cultural, ecological, and organizational revitalization to achieve comprehensive rural revitalization. Column (1) of Table 9 reports the regression with innovation activity (Innovation_ac) as the mediating variable. The results indicate that the coefficient estimate for the explanatory variable is “0.9386” and is statistically significant at the 1% level, demonstrating that AI significantly enhances regional innovation activity. This, in turn, promotes industrial prosperity, ecological livability, rural cultural advancement, effective governance, and affluent living in rural areas. Thus, H3 holds.

Table 9. Mechanism Analysis.

Variable	(1)	(2)
	Innovation Activity	Economic Activity
	Innovation_ac	Economic_ac
AI	0.9386 ***	0.2471 **
	(0.1719)	(0.1166)
Constant	−41.7539 ***	13.0133 **
	(10.2871)	(6.2322)
Control Variable	Yes	Yes
Time effect	Yes	Yes
Individual effect	Yes	Yes
Observed value	403	403
R-squared	0.9452	0.9369

Note: For variable definitions, please refer to Table 2. **, *** correspond to 5%, and 1% significance levels.

4.5.2. Economic Activity

Influenced by location, resource endowments, and existing development patterns, rural industries have long been characterized by limited diversity and low returns, resulting in economic vitality significantly lower than urban areas [10]. With its unique technical advantages of precise matching and intelligent prediction, artificial intelligence not only lowers the technical barriers of digital platforms such as rural e-commerce live streaming systems and smart supply chains but also breaks through the limitations of traditional digital platforms. Through precise supply–demand matching and intelligent resource scheduling, it breaks the physical spatial restrictions of the economy, efficiently integrating scattered rural characteristic products and tourism resources to connect with the national and even global market networks. At the same time, through data-driven precise marketing and business model optimization, it attracts urban capital and consumption power to flow to the countryside more efficiently, giving rise to a large number of new business entities and business models, and significantly stimulating the economic vitality of rural areas [24]. Column (2) in Table 9 reports the regression analysis with economic activity (Economic_ac) as the mediating variable. The coefficient estimates for the explanatory variables are significantly positive, indicating that economic activity plays a significant mediating effect in the process of AI-empowered comprehensive rural revitalization. Therefore, H4 holds.

4.6. Heterogeneity Analysis

4.6.1. Presence of Data Exchange Centers

Data serves as a crucial foundational element for AI model training, pattern discovery, and scenario adaptation. Data trading centers (or exchanges) act as vital channels for data elements to enter the market and further unlock their value. To assess the impact of data marketization on AI’s enabling effects, this study grouped samples based on the presence of data trading centers. The “with data trading centers” group includes 23 provinces/municipalities such as Beijing, Shanghai, Guizhou, Anhui, and Guangdong. Grouped regression analysis was conducted using Model (1). Table 10 presents the results. The results reveal significant heterogeneity in the impact of AI on comprehensive rural revitalization between scenarios with and without data trading centers. The significance level of the explanatory variable coefficient estimates is markedly higher in the group with data trading centers (Column 1) than in the group without (Column 2). This indicates that data trading centers (institutions) provide a superior data element foundation for empowering AI in rural revitalization by aggregating data resources, facilitating data circulation, and promoting the marketization of data elements. They enhance the efficient matching of technology with rural development needs, strengthen the enabling effect of AI on comprehensive rural revitalization, and highlight the crucial role of data element marketization in the synergy between “technology and rural development.”

Table 10. With/Without Data Trading Center.

Variable	(1)	(2)
	With Data Trading Center	No Data Exchange Center
	Rural	Rural
AI	0.0590 **	0.1413 *
	(0.0280)	(0.0728)
Constant	−0.5989	−21.8909 ***
	(1.3470)	(2.3300)
Control Variable	Yes	Yes
Time effect	Yes	Yes
Individual effect	Yes	Yes
Observed value	299	104
R-squared	0.9478	0.9815

Note: For variable definitions, please refer to Table 2. *, **, *** correspond to 10%, 5%, and 1% significance levels.

4.6.2. Population Density Heterogeneity

Artificial intelligence technology exhibits high integration, data dependency, and scenario adaptability. These attributes make population density disparities a significant factor influencing AI-driven comprehensive rural revitalization. To assess the impact of population density differences, this study grouped the sample data based on whether more than 50% of the region lies southeast of the “Hu Huanyong Line” (Heihe-Tengchong Line). The group southeast of the Hu Huanyong Line includes 24 provinces and municipalities—Beijing, Hebei, Shanghai, Guangdong, Sichuan, Jiangsu, etc.—representing China’s most densely populated regions. Conversely, the group northwest of the line encompasses 7 provinces—Qinghai, Inner Mongolia, Gansu, Ningxia, etc.—characterized by relatively sparse population density. Table 11 presents the regression results for these groups. It can be observed that the enabling role of artificial intelligence in comprehensive rural revitalization exhibits significant heterogeneity. In the population-dense region group (Column 1), the coefficient estimate for the explanatory variable is significantly positive. However, in the non-population-dense region group, the coefficient estimate is negative and non-significant, indicating that population density differences are a key factor influencing the enabling effect of artificial intelligence.

Table 11. Regression results of population density heterogeneity grouping.

Variable	(1)	(2)
	Populated Areas (Southeast of Hu Huanyong Line)	Non-Densely Populated Areas (Northwest of the Hu Huanyong Line)
	Rural	Rural
AI	0.0808 ***	−0.0036
	(0.0269)	(0.0856)
Constant	−2.7078 *	−2.1045
	(1.4510)	(3.3171)
Control Variable	Yes	Yes
Time effect	Yes	Yes
Individual effect	Yes	Yes
Observed value	312	91
R-squared	0.9436	0.9806

Note: For variable definitions, please refer to Table 2. *, *** correspond to 10% and 1% significance levels.

4.6.3. Quantile Heterogeneity

To examine the role of artificial intelligence in rural revitalization across different quantile levels, this study established five quantile intervals ranging from the 10th to 90th percentile and conducted regression analysis using a quantile model. Table 12 shows that the enabling effect of AI on comprehensive rural revitalization exhibits significant variation across different quantiles: as the quantile increases, the regression coefficient of AI gradually rises from 0.1459 at the 10th percentile to 0.1887 at the 90th percentile, with all values significant at the 1% level. This result reveals a “quantile gradient effect” in AI’s empowerment of rural revitalization, indicating that initial resource endowments and developmental foundations influence AI’s enabling impact. In lower-quantile rural areas with relatively weaker development bases, while AI still exerts a significant positive effect, its technological efficacy is constrained by factors such as digital infrastructure coverage, talent pool depth, and industrial integration intensity. the release of technological efficacy is less pronounced than in higher-percentage-rank segments. In regions with stronger rural economic foundations and more advanced development levels, AI is more likely to generate synergistic effects, further amplifying technological dividends and producing a Matthew effect where “the strong get stronger.”

Table 12. The results of the quantile regression analysis.

Variable	(1)	(2)	(3)	(4)	(5)
	Rural	Rural	Rural	Rural	Rural
	10th Percentile	25th Percentile	50th Percentile	75th Percentile	90th Percentile
AI	0.1459 ***	0.1485 ***	0.1495 ***	0.1657 ***	0.1887 ***
	(0.0162)	(0.0120)	(0.0172)	(0.0227)	(0.0309)
Constant	−3.5288 ***	−3.2028 ***	−2.6457 ***	−2.4481 ***	−1.3887
	(0.6030)	(0.4483)	(0.6410)	(0.8461)	(1.1511)
Control Variables	Yes	Yes	Yes	Yes	Yes
Observed value	403	403	403	403	403
Pseudo R-squared	0.5585	0.5252	0.4699	0.3942	0.3216

Note: For variable definitions, please refer to Table 2, Since this estimation method does not report the R-squared, the Pseudo R-squared is calculated instead. *** correspond to 1% significance level.

4.6.4. Regional Heterogeneity

Due to various factors such as geographical location and industrial composition, there are significant differences in regional development in China, resulting in a pattern where the eastern region develops industries (Beijing, Liaoning, Tianjin, Zhejiang, Hebei, Shandong, Shanghai, etc., 11 provincial administrative regions), the central region develops agriculture (Shanxi, Hubei, Jilin, Heilongjiang, Henan, Anhui, Jiangxi, Hunan, etc., 8 provincial administrative regions), and the western region develops resource-based industries (Sichuan, Guizhou, Chongqing, Gansu, Yunnan, etc., 12 provincial administrative regions). Among them, the central region has the largest proportion of rural population in China and undertakes the main tasks of grain production and supply and is also the main body of promoting the rural revitalization strategy. Clearly, this pattern of regional economic development and dominant industries will have a significant impact on the enabling effect of artificial intelligence. Therefore, this paper divides the samples into three groups based on the region they are located in and uses model (1) for heterogeneity analysis. The conclusion is shown in Table 13. It can be seen that the enabling effect of artificial intelligence on rural revitalization is most significant in the central region with the highest concentration of rural population, with an estimated coefficient of 0.1404 and significance at the 1% level. In the eastern and western regions, the coefficient estimates of the explanatory variables are negative and not significant, indicating that regional development model differences are important influencing factors of the enabling effect of artificial intelligence.

Table 13. Regression results of regional heterogeneity grouping.

Variable	(1)	(2)	(3)
	Eastern Region	Central Region	Western Region
	Rural	Rural	Rural
AI	−0.0049	0.1404 ***	−0.1441 **
	(0.0623)	(0.0317)	(0.0568)
Constant	−6.4601 ***	3.8691	−14.6526 ***
	(2.0360)	(2.7210)	(2.2000)
Control Variable	Yes	Yes	Yes
Time effect	Yes	Yes	Yes
Individual effect	Yes	Yes	Yes
Observed value	143	104	156
R-squared	0.9445	0.9437	0.9750

Note: For variable definitions, please refer to Table 2. **, *** correspond to 5%, and 1% significance levels.

5. Discussion

5.1. Main Findings

Although China has achieved considerable economic development since the implementation of the “Reform and Opening-up” policy, according to data from the World Bank and the Food and Agriculture Organization of the United Nations, China remains the world’s largest developing country at present, with nearly 2 billion mu of arable land and 500 million rural workers. The income level of rural residents is significantly lower than that of other groups, and rural development and farmers’ income increase are important practical issues that the Chinese government must address at present. Drawing on provincial-level panel data from 2011 to 2023, this study examines the impact of artificial intelligence on comprehensive rural revitalization from both theoretical and empirical perspectives. It explores the patterns and pathways for China’s rural revitalization within the context of disruptive technology integration. The study reveals the following findings: First, artificial intelligence significantly promotes comprehensive rural revitalization by fostering thriving industries, ecological livability, civilized rural customs, effective governance, and prosperous livelihoods in rural areas. This enabling effect exhibits dynamic and sustained characteristics over time. Second, the process of AI empowering rural revitalization is non-linear, exhibiting an inverted U-shaped pattern of initial promotion followed by inhibition. This indicates that excessive technological investment can yield negative effects when institutional development lags behind. Third, enhancing innovation dynamism and economic vitality serves as a crucial transmission pathway for AI’s empowerment of rural revitalization. Fourth, the enabling role of artificial intelligence in the comprehensive revitalization of rural areas will vary significantly depending on whether there is a data trading center, population density, quantile level, and the industrial composition of the region. In samples where there is a data trading center, high population density, high quantile level, and an agricultural-dominated area, the enabling effect of artificial intelligence is more pronounced.

5.2. General Disscussion

Our findings both validate and extend the current literature on AI and rural development. The significant positive effect of AI on comprehensive rural revitalization supports previous work demonstrating AI’s transformative potential in rural contexts [3,4,8,9]. This confirms the theoretical proposition that AI functions as a general-purpose technology applicable across rural production systems [5,14].

However, our study introduces several novel contributions. Unlike existing research that predominantly emphasizes positive outcomes [10,15], we identify an inverted U-shaped relationship with an inflection point at 8.9547. This nonlinear pattern echoes the productivity paradox documented in broader economic contexts [20] but has received limited attention in rural development literature. Our finding challenges prevailing narratives of unqualified technological optimism and demonstrates that excessive AI investment beyond optimal thresholds can yield diminishing or negative returns.

The mechanism analysis reveals that AI empowers rural revitalization through enhancing innovation activity and economic vitality. While these pathways resonate with theoretical frameworks proposed in recent studies [10,24], our research provides rigorous empirical validation through mediation analysis. The stronger effect through innovation channels suggests AI’s primary contribution lies in reshaping rural innovative capacity rather than merely automating existing processes, extending previous work on digital rural governance [16,17].

Our heterogeneity analysis reveals underexplored patterns. The significantly stronger effect in regions with data trading centers underscores the critical role of data infrastructure beyond technical AI capabilities, providing broader empirical support for data-scenario frameworks [18].

5.3. Theoretical Significance

This study advances theoretical understanding through three contributions. First, documenting the inverted U-shaped relationship integrates productivity paradox insights [20] with rural development theory, demonstrating that technological interventions face optimal thresholds beyond which institutional constraints and implementation inefficiencies overwhelm benefits. This suggests rural development strategies must account for absorptive capacity limitations and temporal lags between technology deployment and institutional adaptation.

Second, identifying innovation activity and economic vitality as dual transmission mechanisms provides a granular framework for understanding AI’s systemic influence. The stronger innovation coefficient indicates that AI primarily reduces information asymmetries and enables novel business models—functions aligning with general-purpose technology frameworks [14]. This mechanism-based understanding moves beyond correlational analysis to illuminate causal processes.

Third, dynamic effects analysis demonstrates temporal contingency. The coefficient evolution from 0.1877 (2011) to 0.1501 (2023), with notable intensification during 2018–2020 following national policy implementation, reveals that technological effects depend critically on institutional contexts and policy synchronization.

5.4. Practical Significance

The above conclusions have the following policy implications:

(1) Build smart rural ecosystems to innovate rural development paradigms. Countries should establish national-level smart rural development frameworks, placing AI at the core of rural development strategies. Prioritize digital infrastructure construction, particularly deploying broadband networks and IoT systems in remote areas to lay the groundwork for technology adoption. It is recommended to establish cross-departmental coordination mechanisms, integrating resources from agriculture, technology, education, and other sectors to develop unified technical standards and data protocols. Concurrently, emphasis should be placed on cultivating localized digital solutions, supporting the development of lightweight, low-cost AI tools suitable for smallholder economies to ensure technological inclusivity. By building ecosystems involving multi-stakeholder collaboration among governments, enterprises, and communities, AI can truly become the core force driving systemic transformation in rural areas.

(2) Establish a coordination mechanism between technology governance and institutions, give consideration to enabling efficiency and risk prevention and control, and ensure inclusive development. Although the application of AI technology in rural governance and institutional coordination can improve governance efficiency, it may also generate unique negative problems such as algorithm black box, data rights and interests imbalance, and poor adaptation of governance rules. Without supporting institutional constraints and supervision, it is easy to exacerbate digital exclusion, cause contradictions in interest distribution, and even weaken governance credibility. In order to avoid the diminishing marginal benefits of technology application, countries need to promote technological innovation and institutional innovation simultaneously, establish an adaptive regulatory framework, clarify the rules for the protection of agricultural data rights and interests in rural scenarios, the transparency and interpretability requirements of algorithms, improve the benefit distribution mechanism that takes into account multiple subjects, and pay special attention to the protection of the rights and interests of small farmers and vulnerable groups. Prevent algorithm decision bias or resource misallocation from harming its interests. In addition, financing models should be innovated to combine public and private capital, and the digital divide between urban and rural areas and different business entities should be continuously closed while broadening the scope of technology benefits sharing, so as to prevent the new inequality caused by the application of AI technology from aggravating social contradictions.

(3) Implement differentiated advancement strategies to enhance resource allocation efficiency. Addressing regional development disparities, policy design should adopt a tiered approach. In well-developed areas, prioritize AI applications in advanced scenarios like precision agriculture, supply chain management, and digital governance. In underdeveloped regions, focus on investing in digital infrastructure and human capital, lowering barriers to entry through mobile-first strategies and simplified applications. Crucially, evidence-based policy experimentation mechanisms should be established. Pilot projects in diverse ecological zones will gather localized insights to develop replicable best practices. This incremental, context-specific implementation approach will enhance public resource allocation efficiency and maximize technology-enabled outcomes.

5.5. Limitations and Future Research

This study has some limitations. First, provincial-level data aggregation may obscure significant within-province heterogeneity in AI adoption patterns and rural development outcomes. Second, measuring AI penetration through enterprise counts, while practical, may not fully capture the depth and quality of AI applications in agricultural production and rural governance. Future research could address these limitations through several approaches. Micro-level studies using household data would provide deeper insights into AI adoption mechanisms and distributional impacts across different farmer groups. Cross-national comparative analyses could test whether the inverted U-shaped relationship and enabling mechanisms identified here generalize to other institutional and economic contexts. Additionally, longitudinal case studies tracking specific AI interventions would illuminate the implementation dynamics and long-term sustainability of technology-driven rural transformation.