The Impact of Artificial Intelligence on the Resilience of China’s Fisheries Industry Chain: Evidence from Panel Data Analysis

Abstract

Marine fisheries are a cornerstone of food security, and artificial intelligence (AI) technology holds significant strategic importance for enhancing fishery industrial chain resilience. This study utilizes provincial panel data from China covering the period 2010–2024 to assess the level of fishery industrial chain resilience across 29 provinces (excluding Qinghai, Tibet, Hong Kong, Macao and Taiwan), and employs a two-way fixed-effects model to investigate the mechanisms through which AI influences the resilience of the fisheries industry chain. The findings reveal the following: (1) AI has a significant positive impact on the resilience of the fisheries industry chain. For every one-unit increase in AI development level, the fishery industrial chain resilience increases by 0.470 units; (2) AI enhances fishery industrial chain resilience by promoting structural upgrades in the fisheries industry and improving resource allocation efficiency; (3) heterogeneity analysis indicates that the enabling effects of AI are more pronounced in samples characterized by well-developed digital infrastructure, a higher level of digital economic development, and those located in the eastern regions. This study provides new empirical evidence for understanding how AI empowers the resilience of the fisheries industry chain, and holds significant theoretical and practical value for formulating fisheries science and technology policies tailored to local conditions and for safeguarding national food security.

1. Introduction

As an integral part of agriculture, the fisheries sector plays a pivotal role in safeguarding national food security and implementing the strategy to build a maritime power. The 2025 Central Document No. 1 explicitly calls for ‘enhancing the capacity to ensure the supply of agricultural products’ and ‘establishing a diversified food supply system to promote high-quality development in the fisheries sector’, thereby elevating agricultural food security and fisheries economic development to a new level in which national strategic security and public welfare are advanced in tandem [1]. China is a major agricultural power, as well as the world’s largest producer of fishery products and one of the nations with the richest fishery resources. The healthy development of the marine fisheries economy is not only linked to food security [2] but also constitutes a vital component of implementing the strategy to build a maritime power [1]. However, the fisheries economy is highly dependent on natural resources and is significantly affected by external uncertainties such as climate change, the depletion of aquatic biological resources, and increasing water pollution. Against the backdrop of escalating external risks, reducing the vulnerability of the fisheries economic system and enhancing its resilience to shocks has become a critical issue for food security strategy.

Taking China’s coastal fisheries as an example, offshore fishing in eastern China is concentrated in spring and autumn, while aquaculture in southern China operates year-round yet is highly sensitive to water temperature and water quality. Overfishing has caused a notable decline in some commercial fish stocks, and high-density aquaculture exacerbates water eutrophication and disease risks. Coupled with natural disasters such as typhoons and red tides, as well as fluctuations in the international market, fishery production is confronted with multiple pressures. Enhancing the risk resistance and rapid recovery capacity of the fishery industrial chain has become a strategic issue for safeguarding food security and marine economic stability.

The concept of ‘resilience’ originates from physics, referring to an object’s ability to maintain its equilibrium following an external impact [3], and has since been gradually extended to the field of economics. The 2008 global economic crisis prompted further academic exploration of resilience. Industrial chain resilience stems from the extension and expansion of the concept of ‘resilience’ within the field of industrial economics, referring to the capacity of an industrial chain to recover and reorganize following sudden risks [4]. The resilience of the fisheries industrial chain, meanwhile, refers to the fundamental attribute of the fisheries economic system to maintain its own steady state and, by virtue of sufficient redundancy, return to its pre-disturbance operational trajectory when faced with external sudden shocks and disturbances [5]. With the in-depth implementation of the strategy to build a maritime power, the scale of China’s marine economy has continued to expand, and the marine fisheries sector has entered a phase of rapid development [6]. However, at the same time, a confluence of factors—including frequent marine disasters, the depletion of nearshore resources and an increasingly complex international landscape—poses severe challenges to the stable operation of the marine fisheries economy. Against this backdrop, examining the stable operation of marine fisheries from a resilience perspective holds significant theoretical and practical significance.

At present, AI technology is developing rapidly on a global scale and demonstrating vast prospects for application. As China steadily enters a new phase of development, technological innovation not only serves as the core engine of economic growth, but AI also creates a synergistic effect with the optimization and upgrading of the industrial system, continuously strengthening the economic system’s ability to withstand risks and enhancing economic resilience [7]. In 2022, the Ministry of Science and Technology, together with five other departments, issued the ‘Guiding Opinions on Accelerating Scenario Innovation to Promote High-Quality Economic Development through the High-Level Application of AI’, emphasizing that all regions and relevant entities should accelerate the application of AI in practical scenarios to drive high-quality economic development. To adapt to the rapid development of AI and seize the opportunities presented by the new round of technological revolution, China has elevated the development of AI to a national strategic level. Strong support from national policies has driven the rapid development of AI and its gradual penetration into multiple sectors, including the fisheries industry [8].

In the fishery sector, AI is reshaping production and management across offshore fishing, coastal aquaculture, cooperatives and processing enterprises, covering water quality and fish monitoring, intelligent feeding and oxygenation, fishing route optimization, and digitalization of supply chains. Nevertheless, the fishery industry still faces traditional risks such as resource depletion and natural disasters [9], as well as emerging risks including international market fluctuations and trade barriers. Can the adaptive capabilities afforded by AI technology be effectively translated into the risk resilience of the fisheries industry chain? How exactly does AI influence the resilience of China’s fisheries industry chain? A systematic examination of this positive impact and its underlying mechanisms not only aids in understanding the trajectory of modern fisheries development but also provides a crucial theoretical basis for promoting high-quality development in the sector.

Existing studies have conducted extensive foundational research on AI, industrial chain resilience and fishery intelligent development, which provides important theoretical support for this paper. From the perspective of technological evolution, the application of AI in the fisheries sector can be traced back to the 1980s, with early explorations exemplified by Norway’s intelligent feeding devices and Romania’s expert system for fish disease diagnosis. Relevant research in China began in the 1990s, with representative achievements such as the ‘Fish Doctor’ fish disease diagnosis system developed by Jimei University [10].

Regarding the mechanisms that enhance the resilience of industrial chains, existing research has provided a crucial foundation for this paper from various perspectives. On the one hand, the literature focusing on the fisheries sector confirms that AI can play a positive role in production processes such as disease prevention and control, water quality monitoring, intelligent feeding, and fishery resource management [11,12], and has demonstrated empowerment potential in enhancing production efficiency and upgrading the human capital structure [13], thereby driving the transformation of the fisheries sector toward intelligent operations [14]. On the other hand, research focusing on industrial resilience indicates that digital technologies and AI can enhance the overall risk-resilience and development resilience of the industry by reshaping production models [15,16] and optimizing resource allocation [17], thereby enhancing the sector’s overall resilience to risks and its developmental resilience.

However, existing research still has room for further development in the following areas: firstly, much of the current literature in the fisheries sector focuses on the application of technology in production stages or on efficiency improvements from a single dimension; it has not yet adopted a systematic, full-industry-chain perspective to specifically examine the overall enabling effects of AI on the resilience of the fisheries industry chain; secondly, existing research is predominantly concentrated in the industrial, manufacturing, and broader agricultural and marine economic sectors, lacking detailed analysis of the fisheries industry. The fisheries sector is characterized by significant resource dependence, ecological sensitivity, and production seasonality. The logic underlying the formation of its industrial chain resilience differs fundamentally from that of non-agricultural industries, making it difficult for general resilience theories to effectively explain the unique mechanisms through which AI empowers the fisheries sector; third, although existing literature has addressed the relationship between AI and resilience in agriculture or the marine economy [7], there remains a limited number of studies that specifically target the “Chinese fisheries industry chain” and conduct systematic analyses in conjunction with the characteristics of China’s industrial structure; fourth, the identification of mechanisms is insufficient; existing research lacks systematic empirical testing of the impact of potential mediating pathways—such as industrial structure upgrading, digital transformation, and the optimal allocation of resources—on the resilience of the fisheries industry chain.

Accordingly, this study adopts a holistic perspective on China’s fisheries industry chain, taking 29 provinces (excluding Tibet, Hong Kong, Macao, Taiwan and Qinghai) as its research subjects. It systematically analyses the mechanisms through which AI influences the resilience of the fisheries industry chain and further examines, through empirical testing, regional variations in AI’s enabling effects.

The potential contributions of this paper are as follows: firstly, this study attempts to apply an analytical framework for industrial chain resilience to the field of fisheries economics, constructing an analytical framework based on the dimensions of adaptation and recovery, as well as innovation and restoration. Through empirical testing, it examines the enabling effects of AI on the resilience of the fisheries industrial chain, thereby providing empirical evidence for the role of AI in fisheries development; second, this study empirically reveals how AI enhances the resilience of the fisheries industry chain by driving industrial restructuring and upgrading and improving the efficiency of resource allocation, thereby providing practical empirical insights for understanding the relationship between AI and fisheries development; third, this study further examines the issue of regional disparities in development conditions, investigating the heterogeneous manifestations of AI’s enabling effects across different regions of China, and thus providing targeted empirical evidence to help local governments formulate fisheries development policies tailored to local conditions.

The remainder of this paper is structured as follows: Section 2 covers theoretical analysis and research hypotheses; Section 3 outlines the research design; Section 4 presents the empirical findings; Section 5 examines the mechanisms; and Section 6 concludes with findings and implications.

2. Theoretical Analysis and Research Hypotheses

2.1. AI and the Resilience of the Fisheries Industry Chain

Advances in AI provide a key driver for the development of the fisheries industry chain and play a crucial role in enhancing its ability to cope with multiple uncertainties [18]. Research indicates that when the fisheries industry chain exhibits a high level of AI application, it tends to demonstrate stronger environmental adaptability and dynamic adjustment capabilities [18,19]. Compared with general manufacturing industries and traditional agriculture, fisheries production is characterized by a high degree of “natural dependency” and “spatial mobility,” making its production processes more vulnerable to changes in marine ecosystems, extreme climatic events, disease outbreaks, and international market fluctuations. Consequently, the fisheries sector places higher demands on industrial chain resilience. Leveraging advantages such as real-time sensing, dynamic decision-making, and intelligent coordination, AI technologies provide a novel technological pathway for enhancing the resilience of the fisheries industrial chain.

Firstly, from the perspective of risk resilience, the integration of AI technology directly strengthens the fisheries industry chain’s ability to withstand shocks. Fisheries production is characterized by a pronounced high-risk profile, particularly in aquaculture and offshore fishing activities, where it is frequently exposed to multiple sources of uncertainty, including water quality deterioration, harmful algal blooms, disease outbreaks, and fluctuations in fishery resources. Leveraging AI-driven environmental monitoring sensor networks and water quality prediction and early warning systems, the fisheries sector can effectively mitigate external risks arising from natural disasters and disease outbreaks; simultaneously, through technologies such as big data analytics and intelligent algorithms, fisheries operators are able to accurately forecast market demand and optimize production and fishing plans, thereby mitigating the negative impacts of market fluctuations.

Secondly, according to the theory of optimal resource allocation, the stability of an economic system depends on the rational allocation of production factors [20]. The extensive application of AI technology has enabled the fisheries sector to establish intelligent aquaculture management platforms, allowing precise control over feeding, aeration and water exchange. Furthermore, intelligent algorithms optimize harvesting strategies, promoting the efficient allocation of fishery resources across different production stages and providing robust support for the long-term resilience of the fisheries economic system. Given that fishery resources are characterized by public goods attributes, mobility, and ecological vulnerability, traditional extensive production modes are prone to resource misallocation and overexploitation. AI, through data-driven dynamic adjustment mechanisms, can improve resource utilization efficiency and reduce ecological degradation, thereby enhancing the sustainable supply capacity of the fisheries industrial chain. Additionally, endogenous growth theory emphasizes that technological progress is the endogenous driving force behind economic growth [21]. Within the framework of the Cobb–Douglas production function, AI, as a key driver of technological progress, primarily exerts its influence by enhancing total factor productivity. It not only helps increase unit output in the fisheries sector but also optimizes the allocation efficiency of factors such as capital, labor and resources, thereby driving a shift in the fisheries economic growth model from factor-driven to efficiency-driven, and providing sustained momentum for the sustainable development of the industrial chain.

In summary, the level of AI comprehensively enhances the resilience of the fisheries industry chain in the face of uncertainty by strengthening risk-mitigation capabilities, optimizing the efficiency of resource allocation, and reshaping the growth dynamics. Accordingly, the following hypothesis is proposed:

H1. 

AI has a positive impact on the resilience of the fisheries industry chain.

2.2. The Mediating Role of Industrial Structural Upgrading

Advances in AI are not only a direct driver of increased productivity in the fisheries sector but also a fundamental force behind industrial structural optimization and upgrading, thereby enhancing the resilience of the industrial chain. According to industrial evolution theory, structural upgrading entails the flow of factors from low-productivity sectors to high-productivity sectors, as well as the replacement of traditional production methods by emerging business models [22,23]. For the fisheries sector, industrial structural upgrading is not merely a reallocation of sectoral proportions, but rather an extension of the industrial chain from primary capture fisheries and traditional aquaculture toward higher value-added segments such as advanced processing, smart fisheries, cold-chain logistics, and marine services [24].

From the perspective of production efficiency, AI, as the driving force behind the latest wave of technological transformation, has significantly enhanced the efficiency of fisheries production through technological breakthroughs. For instance, advanced fishing and aquaculture technologies have not only improved production efficiency but also reduced negative environmental impacts, facilitating the transition of the fisheries sector from a traditional model characterized by low efficiency and low value-added to a modern model that is efficient, environmentally friendly and high-value-added. From the perspective of production methods, AI technology, by permeating every link in the fisheries industry chain, has given rise to emerging sectors such as smart equipment manufacturing, intelligent logistics and recreational fisheries, driving the evolution of the industry structure from a traditional, extensive model towards a technology-intensive and high-value-added direction. Consequently, advancements in AI have effectively driven the optimization and upgrading of the fisheries industry structure.

At the same time, this structural upgrade can, on the one hand, extend the length and breadth of the fisheries value chain through the development of value-added processing. Value creation across multiple stages and diversified sources of income can enhance the fisheries sector’s resilience to market and resource fluctuations [23]. On the other hand, by standardizing technology and ensuring information transparency, it eliminates information barriers between different stages of the industrial chain, thereby enhancing industrial interconnectivity and strengthening synergistic linkages among stages, enabling the sector to possess greater self-recovery capacity when facing external shocks [25]. Consequently, AI-driven structural upgrading provides robust technical support and an institutional foundation for the resilience of the fisheries value chain from the perspective of industrial ecosystem evolution. Based on this, the following hypothesis is proposed:

H2. 

AI promotes structural upgrading in the fisheries industry, thereby enhancing the resilience of the fisheries value chain.

2.3. The Mediating Role of Resource Allocation Efficiency

Advances in AI represent a key pathway to improving the efficiency of resource allocation in the fisheries sector and, consequently, strengthening the resilience of the industry chain. From the perspective of resource allocation theory, resource allocation efficiency reflects whether the distribution of production factors among different actors and across various stages of the industry approaches Pareto optimality under given factor endowments [26]. However, the widespread presence of information asymmetry, transaction costs, and institutional frictions makes it difficult to eliminate resource misallocation, significantly constraining the overall efficiency and risk-resilience of the industrial system [27]. Due to characteristics such as long production cycles, dispersed spatial distribution, high market volatility, and incomplete information, resource misallocation is particularly pronounced in the fisheries sector, manifesting primarily as imbalances in the allocation of factors of production across the fishing, aquaculture, processing, and distribution stages.

The introduction of AI provides critical support for alleviating these issues. On the one hand, machine learning and big data analytics can dynamically model catch volumes, marine meteorological conditions, and market prices, significantly improving the accuracy of fishery production planning and market demand forecasting, thereby effectively reducing the misallocation of resources caused by information asymmetry throughout the fishery supply chain [28]. On the other hand, smart sensors and IoT technologies enable end-to-end data visualization across the entire supply chain—from onboard fishing to port processing and final distribution—effectively reducing transaction and coordination costs while improving the efficiency of cross-regional factor flows [29].

Furthermore, improving the efficiency of resource allocation can enhance the resilience of the fisheries industry chain in multiple ways. First, greater allocation efficiency enables the fisheries industry system to rapidly reallocate resources to critical segments in the face of external shocks, effectively shortening the industry chain’s recovery cycle; second, a reduction in resource misallocation helps mitigate the erosion of profits across the entire chain caused by cyclical fluctuations in catch volumes and seafood prices, thereby enhancing the industry’s overall risk resilience; third, an efficient mechanism for factor mobility can strengthen synergies among entities involved in fishing, aquaculture, processing, and distribution, enhancing the fisheries industry system’s long-term adaptability to changes in resource endowments and market structural adjustments [30]. Building on this foundation, AI further amplifies these effects through precise early warning, dynamic scheduling, and intelligent optimization, thereby playing a synergistic role in enhancing the agility and sustainability of the industry chain [28]. In summary, by reducing information asymmetry, optimizing factor flows, and minimizing resource misallocation, AI effectively improves the efficiency of fisheries resource allocation and further enhances the resilience of the fisheries industry chain. Based on this, the following hypothesis is proposed:

H3. 

AI enhances the efficiency of fisheries resource allocation, thereby strengthening the resilience of the fisheries industry chain.

3. Materials and Methods

3.1. Sample Selection and Data Sources

This study utilizes panel data covering the period from 2010 to 2024 for 29 provinces (excluding Tibet, Qinghai, Hong Kong, Macao and Taiwan). The relevant data are sourced from the China Fishery Statistical Yearbook and the China Statistical Yearbook. The original data are generally complete. A small proportion of missing values (approximately 2.06%) is imputed using the linear interpolation method. This approach is suitable for continuous time-series data with scattered and isolated missing observations, as it preserves can maintain the consistency of temporal trends and effectively avoid sample loss.

3.2. Definition and Description of Variables

3.2.1. Dependent Variable

Industrial Chain Resilience (ICR) is the dependent variable in this study. Industrial chain resilience refers to the capability of an industrial chain system to maintain normal operation, achieve rapid recovery, and realize continuous optimization and upgrading when confronted with uncertain factors such as external shocks, market fluctuations, and natural disasters. Its core connotation covers multiple dimensions including resistance capability, recovery capability, adaptability, and dynamic evolution capability [31]. Compared with general industrial chains, the fisheries industrial chain is more susceptible to natural resource constraints, ecological environment changes and market risks. Therefore, its resilience is reflected not only in the short-term ability to withstand shocks, but also in its long-term adjustment and sustainable development capacities.

Based on the discussion and drawing on Wu et al. [32] and Holzhacker et al. [31], this paper constructs an evaluation indicator system for the resilience of the fisheries industrial chain from four dimensions: recovery capacity, adaptive capacity, innovation capacity, and restoration capacity. Among them, recovery capacity mainly reflects the ability of the fisheries industrial chain to sustain production and economic recovery after external shocks. Adaptive capacity reflects the industrial chain’s ability to allocate resources and adjust production in response to changes in the external environment. Innovation capacity reflects the level of technological progress, factor allocation optimization, and industrial upgrading within the industrial chain. Restoration capacity represents the recovery and improvement capacity of the fisheries industrial chain in terms of ecological governance and green sustainable development.

Finally, the entropy method is adopted to calculate the weight of each dimension and the comprehensive resilience score, as shown in

Table 1. Indicator system for the resilience of the fisheries industry chain.

Primary Indicator	Secondary Indicators	Third-Level Indicators
Resilience of the fisheries industry chain	Recovery capacity	Total output value of the fisheries sector
		Per capita net income of fishermen nationwide
		Aquaculture area
		Direct economic losses caused by the disaster
	Adaptability capacity	fisheries workers
		Total fish production
		Fishing vessel fleet at year-end
	Innovation capacity	Number of aquaculture extension agencies
		Actual number of staff involved in aquaculture technology extension
		Total funding for aquaculture extension agencies
		Value of imports and exports of aquatic products
	Restoration capacity	Local government expenditure on environmental protection
		Volume of industrial wastewater discharged

.

3.2.2. Explanatory Variables

The core explanatory variable in this study is the level of AI in each region. As machine learning and deep learning form the core of AI technology [33], this study utilizes patent search and statistical data from the China National Intellectual Property Administration (CNIPA) to identify AI patent applications and grants using the keywords ‘machine learning’ and ‘deep learning’. AI patents serve as one of the direct indicators of the level of AI development. Drawing on the research of Fujii and Managi [34] and Ding and Gao [35], this study adopts the number of authorized AI patents in each province as the core explanatory variable. The scale of regional AI patent grants reflects a region’s technological R&D capabilities and industrial application foundation, and also indicates the fisheries industry’s potential capacity and enabling conditions for implementing AI technology; therefore, this study uses it as a proxy variable to measure the level of AI application.

3.2.3. Control Variables

Drawing on existing literature [32,36], this study selects the following control variables: Gross Output Value of Fishery (GOV), a core indicator characterizing the scale and volume of the fisheries economy; Added Value of Fishery (AVF), measured by the gross output value of the fishery minus the value of intermediate products consumed during the production process; Degree of Opening Up (OPEN), measured by the ratio of total local imports and exports to local gross domestic product; Fiscal Support Intensity (FSI), measured by the ratio of general budget expenditure from local government to local GDP; Urbanization level (URB) is expressed as the ratio of the permanent resident population at year-end to the urban population; Fiscal expenditure on science and technology (FEST) is measured by actual local government expenditure in the science and technology sector; Direct economic loss from natural disasters (LOSS) is expressed as the monetary value of direct losses caused by natural disasters. Furthermore, this study controls for individual and year fixed effects.

3.3. Model Specification

This paper empirically examines the impact of AI on the resilience of China’s fisheries industry chain by constructing a panel data model. The baseline model is specified as follows:

$$\mathrm{I}{\mathrm{C}\mathrm{R}}_{\mathrm{i}\mathrm{,}\mathrm{t}}=\mathsf{\alpha }+\mathsf{\beta }{\mathrm{A}\mathrm{I}}_{\mathrm{i}\mathrm{,}\mathrm{t}}+\mathsf{\gamma }{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i}\mathrm{,}\mathrm{t}}+{\mathrm{\mu }}_{\mathrm{i}}\mathrm{+}{\mathrm{\phi }}_{\mathrm{t}}\mathrm{+}{\mathrm{\epsilon }}_{\mathrm{i}\mathrm{,}\mathrm{t}}$$

(1)

where ${ICR}_{i,t}$ represents the level of resilience of China’s fisheries industry chain, ${AI}_{i,t}$ represents the provincial level of AI, and ${Controls}_{i,t}$ represents control variables at the provincial level; ${\mu }_i$ and ${\phi }_t$ represent provincial fixed effects and time fixed effects, respectively, used to control for potential influencing factors that do not vary across provinces or over time; ${\epsilon }_{i,t}$ is the random error term.

4. Results

4.1. Descriptive Statistics

The results of the descriptive statistics for the core variables and control variables are shown in

Table 2. Results of descriptive statistics.

Variables	Mean	Std	Min	P25	P50	P75	Max
ICR	1,136,473.00	1,397,826.00	12,176.07	172,404.10	496,442.10	1,609,951.00	7,481,856.00
AI	6416.12	12,519.14	10.00	705.00	2255.00	6299.00	92,987.00
GOV	4.00	5.12	0.01	0.35	1.42	5.76	20.98
AVF	0.91	1.89	0.00	0.00	0.00	0.49	9.36
OPEN	0.42	0.42	0.04	0.15	0.23	0.54	2.21
FSI	0.24	0.08	0.10	0.18	0.22	0.29	0.45
URB	1.72	0.35	1.11	1.49	1.67	1.91	2.96
FEST	1.57	1.93	0.06	0.41	0.76	2.04	11.69
LOSS	5691.19	3290.96	557.53	3369.18	5011.56	7205.12	18,533.08

. The distribution characteristics and statistical patterns of the sample data provide a foundation for the subsequent empirical analysis. The mean value of the fisheries industry chain resilience (ICR) is 1,136,473.00, with a standard deviation of 1,397,826.00, a minimum value of 12,176.07 and a maximum value of 74,818,360.00. This indicates that there are significant differences in the levels of industry chain resilience across the sample, and that the distribution exhibits a right-skewed characteristic. The mean value of the AI level is 6416.12, with a standard deviation of 12,519.14, a minimum value of 10.00, a maximum value of 92,987.00, and a median of 2255.00, similarly exhibiting distinct regional differentiation and a right-skewed distribution. Overall, both core variables exhibit strong dispersion, with the majority of samples concentrated in the lower range, whilst a small number of high-value samples have elevated the mean. Some variables exhibit high levels of dispersion and skewed distributions. In the baseline regression, this study employs robust standard errors to mitigate estimation biases caused by non-normal distributions. Additionally, log transformations were applied to key variables to conduct robustness tests, and the results indicate that the study’s conclusions are robust. The results of the multicollinearity test show that all variables have variance inflation factor (VIF) values below the conventional academic threshold of 10, with an average VIF of 3.34. This indicates that there is no severe multicollinearity issue between the variables, which supports subsequent empirical analysis.

The results of the correlation analysis in

Table 3. Variables of correlation coefficients.

Variables	(1) ICR	(2) AI	(3) GOV	(4) AVF	(5) OPEN	(6) FSI	(7) URB	(8) FEST
(2) AI	0.574 ***	1.000
(3) GOV	0.941 ***	0.489 ***	1.000
(4) AVF	0.275 ***	−0.069	0.356 ***	1.000
(5) OPEN	0.196 ***	0.412 ***	0.219 ***	0.213 ***	1.000
(6) FSI	−0.597 ***	−0.412 ***	−0.630 ***	−0.354 ***	−0.441 ***	1.000
(7) URB	−0.262 ***	−0.432 ***	−0.271 ***	−0.057	−0.607 ***	0.370 ***	1.000
(8) FEST	0.674 ***	0.849 ***	0.615 ***	−0.006	0.461 ***	−0.526 ***	−0.472 ***	1.000
(9) LOSS	0.689 ***	0.722 ***	0.629 ***	−0.022	0.203 ***	−0.518 ***	0.367 ***	0.842 ***

*** indicates significance at the 1% level (p < 0.01).

indicate that there is a significant positive correlation between the level of AI and the resilience of the fisheries industry chain, with a correlation coefficient of 0.574, which is statistically significant at the 1% level (p < 0.01). This finding suggests that improvements in the level of AI are closely associated with enhanced resilience in the fisheries industry chain, thereby providing preliminary validation of the core research hypothesis of this study.

4.2. Results of the Baseline Regression

This paper analyses the impact of AI on the resilience of China’s fisheries industry chain by constructing a multiple regression model. The results of the baseline regression are shown in

Table 4. Benchmark regression results.

Variables	(1)	(2)	(3)
	ICR	ICR	ICR
AI	0.470 ***	0.370 ***	0.175 ***
	(0.023)	(0.025)	(0.031)
GOV			0.520 ***
			(0.058)
AVF			−0.070 ***
			(0.021)
OPEN			0.032
			(0.054)
FSI			−0.055
			(0.049)
URB			−0.069
			(0.057)
FEST			−0.013
			(0.045)
LOSS			0.247 ***
			(0.070)
Individual fixed	NO	YES	YES
Year fixed	NO	YES	YES
Observations	435	435	435
R2		0.602	0.772

Standard errors in parentheses; *** p < 0.01.

Column (1) presents the regression results without fixed effects; the coefficient for AI is 0.470 and is significant at the 1% level. In Column (2), after incorporating individual and year fixed effects, the coefficient for AI is 0.370 and remains significant at the 1% level; In Column (3), after further incorporating all control variables and controlling for individual and year fixed effects, the coefficient for AI is 0.175, which is significantly positive at the 1% level. This indicates that, regardless of whether fixed effects and control variables are controlled for, AI exerts a significant positive impact on the resilience of the fisheries industry chain, and the core conclusion is reliable.

From an economic perspective, after controlling for other influencing factors, a one-unit increase in AI level raises the comprehensive score of fishery industrial chain resilience by 0.175. This effect carries important practical implications for the fishery sector: AI application helps optimize fishery production scheduling, improve supply chain information transparency, and reduce losses caused by natural risk shocks, thereby systematically enhancing the resistance and recovery capacity of the industrial chain.

4.3. Robustness Tests

4.3.1. Replacing Explanatory Variables

This study selected sub-dimensions of the core explanatory variables to re-measure the level of AI and conducted separate regressions. The indicators for the number of AI patents obtained include invention patents obtained (INO) and utility model patents obtained (PRA). See

Table 5. Results of the robustness test.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	ICR	ICR	ICR	ICR	ICR	ICR	ICR	ICR
INV	0.093 ***
	(0.024)
PRA	0.226 ***
	(0.035)
L.AI					0.021 ***
					(0.006)
AI		0.011 ***	0.058 ***	0.092 ***		0.230 ***	0.118 ***	0.134 ***
		(0.004)	(0.012)	(0.027)		(0.069)	(0.036)	(0.031)
GOV	0.538 ***	0.571 ***	0.007	0.015	0.525 ***	0.240 ***	0.504 ***	0.515 ***
	(0.059)	(0.101)	(0.009)	(0.049)	(0.057)	(0.045)	(0.054)	(0.058)
AVF	−0.075 ***	−0.071 ***	0.007 **	0.045 **	−0.049 **	0.016	−0.083 ***	−0.060 ***
	(0.022)	(0.024)	(0.003)	(0.018)	(0.021)	(0.017)	(0.019)	(0.021)
OPEN	−0.011	−0.067	−0.032 ***	0.048	0.004	0.096 **	0.043	0.042
	(0.054)	(0.091)	(0.008)	(0.046)	(0.052)	(0.040)	(0.049)	(0.055)
FSI	−0.069	−0.067	0.000	−0.050	−0.052	−0.033	−0.055	−0.075
	(0.050)	(0.046)	(0.008)	(0.042)	(0.048)	(0.039)	(0.044)	(0.051)
URB	−0.052	−0.024 ***	−0.037 ***	−0.059	−0.041	−0.024	−0.043	−0.049
	(0.058)	(0.046)	(0.010)	(0.049)	(0.056)	(0.046)	(0.052)	(0.059)
FEST	0.007	−0.032	−0.022 ***	−0.014	−0.010	0.119 ***	−0.013	0.017
	(0.045)	(0.072)	(0.007)	(0.038)	(0.044)	(0.035)	(0.046)	(0.044)
LOSS	0.304 ***	0.016	0.015	0.157 ***	0.186 ***	0.003	0.216 ***	0.196 ***
	(0.070)	(0.065)	(0.012)	(0.060)	(0.071)	(0.056)	(0.066)	(0.072)
Individual/Year fixed	YES	YES	YES	YES	YES	YES	YES	YES
Observations	435	435	435	435	435	435	435	377
R2	0.762	0.757	0.313	0.713	0.777	0.632	0.749	0.763

Standard errors in parentheses; ** p < 0.05, *** p < 0.01.

. The regression results in Column (1) show that the coefficients for both INO and PRA are significantly positive at the 1% level, indicating that AI significantly promotes the resilience of the fisheries industry chain, and the core conclusions remain robust.

Furthermore, to ensure the robustness of the findings, this study replaces the measurement of the explanatory variable and re-estimates the level of AI application across provinces. To scientifically measure provincial AI development, a keyword frequency analysis was conducted based on provincial government work reports. Since these reports are generally released at the beginning of each year, their content does not possess predictive power regarding actual economic performance within the same year; however, they can, to some extent, reflect local governments’ emphasis on AI development. Using Python 3.13-based text mining techniques, this study calculates the frequencies of keywords related to AI applications, including “AI,” “intelligence,” “intelligentization,” “robots,” “cloud computing,” “smart,” “smart agriculture,” “big data,” “digitalization,” “human-computer interaction,” and “Internet of Things.” To avoid measurement bias caused by differences in text length, the ratio of keyword frequency to total word frequency is adopted as an alternative measure, thereby enhancing the comparability and robustness of the variable measurement. The regression results reported in Column (2) show that the coefficient of AI is significantly positive at the 1% level, indicating that AI significantly enhances the resilience of the fishery industrial chain, thereby confirming the robustness of the core findings.

4.3.2. Replacement of the Explained Variable

In the benchmark regression, the entropy method is adopted to quantify the resilience level of each sample. To avoid the one-sidedness and measurement errors caused by a single measurement method, this paper replaces the calculation approach of the explained variable, and re-measures the resilience level of the fisheries industrial chain using the entropy-weighted TOPSIS method for regression analysis. As shown in Column (3), the coefficient of AI is significantly positive at the 1% statistical level, indicating that AI can significantly enhance the resilience of the fisheries industrial chain, and the core research conclusion remains robust.

In addition, the foregoing descriptive statistics indicate that the dependent variable exhibits high dispersion and skewed distribution, which may exert potential interference on the baseline regression estimation results. To further verify the reliability of the research conclusions, this paper performs a logarithmic transformation of the dependent variable and re-estimates the regression model. Logarithmic transformation can effectively smooth the data distribution, weaken the impact of extreme values and right-skewness, and mitigate estimation bias caused by high dispersion. As shown in Column (4), the coefficient of AI remains significantly positive at the 1% statistical level, confirming the robustness of our core findings.

4.3.3. One-Period Lags of Explanatory Variables

The mutual influence between the dependent variable and the independent variables may lead to endogeneity issues; applying a one-period lag to the explanatory variables helps mitigate the endogeneity interference caused by reverse causality [37]. This paper applies a one-period lag to the explanatory variables to conduct robustness tests. The regression results in Column (5) show that the coefficient for AI is significantly positive at the 1% level, consistent with the previous regression results, and the core conclusion remains robust.

4.3.4. Truncation of Variables

To mitigate the potential impact of biases in the collection and measurement of macroeconomic variables on the study’s conclusions, and drawing on the research by Shao Shuai et al. [38], this study applied a trimming procedure to the sample data of the core variables, removing the top and bottom 10% of values. Parameter estimates were then recalculated based on the trimmed data, with the corresponding results presented in the table below. The results in Column (6) indicate that, following the trimming procedure, the study’s baseline conclusions remain robust.

4.3.5. Adjustment of Sample Size

Given that natural disasters such as typhoons and red tides may interfere with the impact of AI on the resilience of the fisheries industry chain, this study excluded two provinces from the 29-province sample where such disasters occur frequently: Hainan Province and Guangdong Province [32]. The regression results in Column (7) indicate that the AI coefficient is significantly positive at the 1% level, and the core conclusions of this paper remain robust.

4.3.6. Excluding the Impact of the Pandemic Years

To rule out the extreme disruptive impacts of the COVID-19 pandemic on fishery industrial chain operations, this study re-estimates the baseline model by excluding the sample observations from 2021 to 2022. As shown in Column (8), the coefficient of AI is significantly positive at the 1% statistical level. This indicates that the promotional effect of AI on the resilience of China’s fishery industrial chain is not driven by the special impact of the COVID-19 pandemic, but reflects a stable effect under normal economic conditions, further verifying the robustness and reliability of the research conclusions.

4.4. Heterogeneity Analysis

4.4.1. Heterogeneity in Digital Infrastructure

A well-developed digital infrastructure not only supports the real-time collection and efficient circulation of massive amounts of data, but also expands the scope for integration between AI and the fisheries industry chain, accelerating the implementation of intelligent technologies in fisheries production, processing and distribution [39]. Consequently, the impact of AI on the resilience of the fisheries industry chain may exhibit heterogeneity at the level of digital infrastructure.

Digital infrastructure is a multi-faceted and multi-level systematic project, and a single indicator cannot comprehensively measure its true development level. Considering the availability of data at the provincial level and referring to the studies by Yin et al. [40] and Tang and Yany [41], this paper constructs an evaluation indicator system for digital infrastructure from two dimensions: digital infrastructure input and digital infrastructure output. For the input dimension, three indicators are selected to measure the construction foundation and resource input intensity: optical cable density (length of long-distance optical cable lines divided by administrative area), per capita Internet broadband access ports (number of Internet broadband access ports divided by total population), and the proportion of urban employees in the information transmission, computer services, and software industries. For the output dimension, three indicators are chosen to reflect application effectiveness and popularization level: per capita telecommunications business income (total telecommunications business income divided by total population), mobile phone penetration rate (number of mobile phone users divided by total population), and Internet penetration rate (number of Internet broadband access users divided by total population). All indicators are positive indicators, meaning higher values represent a better development level of digital infrastructure.

When conducting grouped regression using the median level of digital infrastructure as the cutoff point, the results in

Table 6. Results of the heterogeneity analysis.

	Resilience of the Fisheries Industry Chain (ICR)
	(1) High Infrastructure	(2) Low Infrastructure	(3) High Digital Economy	(4) Low Digital Economy	(5) Eastern	(6) Central and Western
AI	0.177 ***	0.117 **	0.220 ***	0.099	0.228 ***	0.088
	(0.059)	(0.046)	(0.045)	(0.400)	(0.041)	(0.135)
GOV	0.562 ***	0.471 ***	0.414 ***	0.236 *	0.427 ***	0.614 ***
	(0.149)	(0.060)	(0.105)	(0.137)	(0.086)	(0.094)
AVF	−0.010	−0.118 ***	−0.052	−0.082 *	−0.085 ***	−0.079 *
	(0.045)	(0.025)	(0.039)	(0.042)	(0.031)	(0.042)
OPEN	0.015	0.016	−0.036	0.012	−0.027	−0.100
	(0.154)	(0.058)	(0.109)	(0.107)	(0.099)	(0.116)
FSI	−0.167	−0.088 *	−0.302 **	−0.091	−0.350 ***	−0.045
	(0.158)	(0.049)	(0.147)	(0.058)	(0.130)	(0.048)
URB	−0.093	−0.006	−0.644 *	0.154 *	−0.402 **	0.047
	(0.201)	(0.057)	(0.338)	(0.079)	(0.188)	(0.062)
FEST	−0.033	−0.135 **	−0.111	0.389 ***	−0.066	0.035
	(0.085)	(0.066)	(0.075)	(0.146)	(0.075)	(0.059)
LOSS	0.554 ***	0.289 ***	0.438 ***	0.042	0.373 ***	0.167 *
	(0.194)	(0.086)	(0.155)	(0.133)	(0.128)	(0.902)
Individual/Year	YES	YES	YES	YES	YES	YES
Observations	208	227	217	218	165	270
R2	0.687	0.780	0.763	0.416	0.867	0.685

Standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

show that the coefficient for AI in the high digital infrastructure group is 0.177 and is significantly positive at the 1% level, whilst the coefficient for AI in the low digital infrastructure group is 0.117 and is significantly positive at the 5% level. This indicates that, regardless of the level of digital infrastructure, AI has a positive impact on the resilience of the fisheries industry chain, with the effect being stronger in the high digital infrastructure group. A possible explanation is that well-developed infrastructure breaks down data barriers between different links in the industrial chain, enabling AI to coordinate the allocation of emergency resources such as cold storage and transport capacity, thereby enhancing the overall collaborative risk-resilience of the industrial chain.

4.4.2. Heterogeneity in the Level of Digital Economy Development

There are significant disparities in the level of digital economic development across China’s provinces; some provinces are relatively advanced in terms of technological innovation capacity and digital transformation, whilst others lag. As the digital economy is closely linked to the application of AI, the impact of AI on the resilience of the fisheries industry chain may also exhibit heterogeneity across different levels of digital economic development. Drawing on the research by Zhao Tao et al. [42], we measure the comprehensive level of digital economic development from two perspectives: internet development and digital financial inclusion. Firstly, drawing on the methodology of Huang et al. [43], we employ four indicators: internet penetration, employment in related sectors, related output, and mobile phone penetration. These indicators are measured respectively by the number of broadband internet users per 100 people, the proportion of employees in the computer services and software industry relative to total urban employed personnel, the total volume of telecommunications services per capita, and the number of mobile phone users per 100 people. Data sources are drawn from the “China Urban Statistical Yearbook”. Regarding the development of digital finance, the China Digital Inclusive Finance Index is used as a measure; this index is jointly compiled by the Digital Finance Research Centre at Peking University and Ant Financial Group [44].

When conducting grouped regression using the median of the digital economy index as the cutoff point, the results show that the coefficient for AI in the high-digital-economy group is significantly positive at the 1% level, whereas the coefficient for AI in the low-digital-economy group is not statistically significant. This suggests that the effect of AI on enhancing the resilience of the fisheries industry chain may be subject to a certain ‘digital economy threshold’. In the high digital economy group, the relatively active integration of digital technologies with the real economy has created a more mature market environment and industrial infrastructure for the application of AI technology in fisheries production, processing and distribution. At the same time, such regions typically possess a stronger talent pool, which helps fishery practitioners to adopt and apply AI technology. In contrast, in the low digital economy group, industrial digital transformation is still in its infancy; information connectivity across various links in the industrial chain needs to be strengthened, and the depth and breadth of technology application are relatively limited, meaning that the positive impact of AI technology on industrial chain resilience has yet to be fully realized.

4.4.3. Regional Heterogeneity

Due to differences in resource endowments and stages of development, there is marked heterogeneity in the regional distribution of AI technology adoption. Consequently, the impact of AI on the resilience of the fisheries industry chain may also vary across regions and provinces, and it is necessary to examine this in greater depth.

Based on geographical differences, this study divided the sample into eastern and central-western regions for separate regression analyses. The results show that in the eastern region group, the coefficient for AI was 0.228 and was significantly positive at the 1% level, whereas in the central-western region group, the coefficient for AI was not statistically significant. This may be attributed to the fact that, as a frontier of opening to the outside world and a hub of consumer markets, the eastern region’s fisheries industry is deeply embedded in global supply chains. It faces more frequent fluctuations in international markets and stricter quality standards; such external pressures compel business operators to proactively utilize AI to enhance their responsiveness and risk management capabilities. In contrast, in the central and western regions, fisheries production primarily serves regional domestic markets. The industrial chain is relatively closed, with stable supply-demand dynamics and a lack of external shocks to drive technological upgrading; consequently, the role of AI in enhancing the resilience of the industrial chain has not yet been fully realized.

5. Mechanism Testing

5.1. Transformation and Upgrading of the Industrial Structure

The results of the benchmark regression confirm that technological innovation has a significant positive impact on the resilience of China’s fisheries industry chain. Drawing on the approaches adopted by Jiang and Ting [45] and Chen et al. [46], this study further examines the transmission mechanisms to reveal the underlying logic by which AI enhances the resilience of the fisheries industry chain.

This paper aims to elucidate the mechanism through which AI enhances the resilience of the fisheries industry chain by examining the pathway of industrial structure transformation and upgrading. In research on industrial structure upgrading, the academic community has primarily focused on two dimensions: industrial structure upgrading and rationalization. Industrial structure upgrading, also known as industrial advancement, is manifested in the dynamic evolution of the focus of the three sectors of the economy, typically exhibiting a progressive shift from labor-intensive to capital-intensive industries [47]. The rationalization of industrial structure, on the other hand, focuses on the degree of coordination between the efficiency of factor allocation and industrial output; the more optimized the combination of factors and the more balanced the synergy between industries, the higher the level of rationalization. Consequently, this study measures the transformation and upgrading of the industrial structure through the dual dimensions of the rationalization of the industrial structure (RIS) and the upgrading of the industrial structure (UIS).

Firstly, drawing on relevant research [48], this paper employs an industrial structure deviation index based on the Theil index to measure the level of rationalization of the industrial structure in each city. Industrial structure rationalization is primarily used to reflect the degree of coordination between industries and the efficiency of resource allocation. A rational industrial structure can effectively mitigate the vulnerability arising from the fisheries economy’s over-reliance on a single industry, strengthen the sector’s resilience to external economic fluctuations, and thereby have a positive impact on the resilience of the industrial chain. The specific calculation method is as follows:

$$RIS={\displaystyle {\sum }_{i=1}^3}(\frac{Y_i}{Y})\mathrm{l}\mathrm{n}(\frac{Y_i}{L_i}/\frac{Y}{L})$$

(2)

Here, $Y$ represents the regional gross domestic product across the three sectors, whilst $Y_i$ and $L_i$ denote the output value and employment figures for sector i, respectively. When all sectors are in a state of balanced development, the Thiel Index takes the value of 0. A higher index value indicates a tendency towards a monolithic industrial composition and a lower degree of rationalisation in the industrial structure; conversely, a lower index value indicates a higher level of rationalisation in the industrial structure.

Secondly, to measure the level of industrial structure upgrading, this paper adopts the ratio of the output value of the tertiary sector to that of the secondary sector as a proxy indicator (UIS). This indicator effectively captures the trend of the economic structure’s transition towards a service-oriented economy and clearly reflects whether the industrial structure exhibits ‘service-oriented’ characteristics; it is therefore considered to possess good measurement validity in relevant research [48]. If the UIS value tends to rise, this indicates that the economic system is gradually moving towards a service-oriented direction, and that the industrial structure is undergoing an upgrade.

As shown in

Table 7. Results of the mechanism test.

Variables	(1)	(2)	(3)	(4)
	RIS	UIS	DCM	DLM
AI	−0.106 ***	0.059 ***	−0.022 ***	−0.100 ***
	(0.036)	(0.016)	(0.008)	(0.033)
GOV	0.283 ***	−0.087 **	−0.093 ***	0.005
	(0.074)	(0.039)	(0.020)	(0.065)
AVF	0.024	0.023 *	−0.016 ***	0.016
	(0.021)	(0.013)	(0.004)	(0.010)
OPEN	−0.175 **	−0.161 ***	0.016	0.053
	(0.087)	(0.041)	(0.013)	(0.059)
FSI	0.085	−0.265 ***	−0.013	−0.030
	(0.077)	(0.047)	(0.015)	(0.040)
URB	−1.285 ***	0.023	0.146 ***	0.151 *
	(0.115)	(0.058)	(0.021)	(0.079)
FEST	0.143 **	−0.049	−0.067 ***	−0.012
	(0.060)	(0.036)	(0.011)	(0.046)
LOSS	−0.599 ***	−0.017	0.102 ***	0.135 **
	(0.105)	(0.069)	(0.020)	(0.066)
Individual/Year	YES	YES	YES	YES
Observations	435	435	435	435
R2	0.913	0.975	0.980	0.866

Standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

, the regression results indicate that the coefficient for RIS is −0.106 and is significantly negative at the 1% level, whilst the coefficient for UIS is 0.059 and is significantly positive at the 1% level; furthermore, the impact of AI on the rationalization of industrial structure is more pronounced. This suggests that AI has a significant positive impact on the rationalization and upgrading of the fisheries industry structure, thereby enhancing the resilience of the fisheries industry chain; H2 is thus confirmed.

5.2. Efficiency of Resource Allocation

The theoretical analysis in this paper demonstrates that AI applications can enhance the resilience of the fisheries industry chain by optimizing the efficiency of resource allocation. Effective resource allocation refers to a state achieved under market mechanisms where the free flow of factors maximizes economic efficiency and optimizes social benefits, whereas distortions in factor markets signify a deviation from this state. Drawing on the research by Liu and Xia [49], this paper employs the production function method to measure the degree of distortion in the capital and labor markets. The calculation process is as follows.

First, a Cobb–Douglas production function (C-D production function) is specified and taken to the natural logarithm.

$$\ln Y_{it}=\mathrm{c}+\alpha \ln K_{it},+\beta \ln L_{it}.+\epsilon$$

(3)

In this model, Y is estimated using regional gross domestic product (GDP); K represents the capital stock, estimated using the perpetual inventory method; L represents the labor force, expressed as the number of employed persons in each city at the end of the year; and the marginal products of capital and labor are $\alpha Y_{it}$/$K_{it}$, $\beta Y_{it}$=/$L_{it}$, respectively.

Second, by measuring the degree of market distortion based on the deviation between the marginal product of a factor and its price, the degrees of distortion in the capital market and the labor market are respectively:

$${DCM}_{it}=|\raisebox{1ex}{$\alpha Y_{it}$}\!\left/ \!\raisebox{-1ex}{$r{K}_{it}$}\right.-1|$$

(4)

$${DLM}_{it}=|\raisebox{1ex}{$\beta Y_{it}$}\!\left/ \!\raisebox{-1ex}{$w{L}_{it}$}\right.-1|$$

(5)

Here, r represents the price of capital, set at 10% to reflect a 5% depreciation rate and a 5% real interest rate; w represents the price of labor, expressed as the average wage of employed persons in each city for that year; and the capital-output elasticity α and labor-output elasticity β are obtained through regression based on the C-D production function. Therefore, the smaller the values of DCM and DLM, the lower the degree of distortion in capital and labor, and the higher the efficiency of their allocation.

The regression results are shown in Table 7, Column (3) indicates that the coefficient for DCM is −0.022 and is significantly negative at the 1% level, while the coefficient for DLM is −0.100 and is significantly negative at the 1% level; furthermore, AI has a more significant impact on labor resource allocation. This suggests that AI can effectively reduce the degree of resource misallocation across regions, thereby enhancing the resilience of the fisheries industry chain, thus confirming H3.

6. Discussion

Based on provincial-level panel data from China covering the period 2010–2024, this paper first estimates the resilience of the fisheries industry chain and then employs a two-way fixed-effects model to investigate the impact and mechanisms of AI on the resilience of the fisheries industry chain. The main findings are as follows:

Firstly, AI has a significant positive impact on the resilience of the fisheries industry chain. This conclusion confirms that AI can also play a role in risk mitigation and recovery enhancement within the fisheries sector, consistent with existing research findings in fields such as manufacturing and regional economics [50,51]. Distinguishing itself from previous analyses that have largely focused on industrial or macro-level perspectives [52], this study extends the relationship between AI and industrial chain resilience to the specific sector of marine fisheries, confirming that even within fisheries systems characterized by high resource dependency and ecological sensitivity, AI technology remains a key driver of resilience enhancement under external shocks. Furthermore, the role of AI in enhancing the resilience of the fisheries industry chain is not limited to fish stock monitoring [53,54], but extends widely to key areas such as environmental monitoring and sustainable fisheries development [55,56], thereby providing multi-dimensional support for the high-quality development of marine fisheries. This finding provides new empirical evidence for understanding the modernization of fisheries and expands the scope of application of industrial chain resilience theory within the agricultural sector.

Secondly, the upgrading of the fisheries industry structure serves as a key channel through which AI enhances the resilience of the fisheries industrial chain. Existing research indicates that the level of economic development and labor productivity in the primary fisheries sector are key factors influencing the upgrading of China’s fisheries industry [57]. As a major driver of labor productivity and economic growth, AI can provide strong support for enhancing the resilience of the fisheries industry chain by promoting the upgrading of the industry’s structure; simultaneously, upgrading, advancement and rationalization of the industrial structure are also regarded as direct drivers of economic resilience [24]. While existing literature largely treats the above mechanisms as mutually independent processes, this study adopts a mediating effects perspective to empirically examine the role of fisheries industrial restructuring in mediating the relationship between AI and industrial chain resilience, thereby providing a new analytical framework for understanding the intrinsic logic among the ‘technology–structure–resilience’ triad.

Thirdly, the impact of AI on enhancing the resilience of the fisheries industry chain exhibits regional heterogeneity and is conditional on specific circumstances. An analysis of this heterogeneity indicates that the resilience-enhancing effects of AI are more pronounced in samples characterized by relatively well-developed digital infrastructure, a higher level of digital economic development, and those located in the eastern regions. This finding is consistent with the conclusion drawn in certain regional economic studies that ‘innovation agglomeration effects are more favorable to developed regions’ [58]. This is because regions with well-developed digital infrastructure and a mature digital economy can provide more abundant data resources for the implementation of AI technologies. Furthermore, the eastern regions, with their established industrial chain support systems, are better positioned to absorb and transform the technological dividends brought by AI, thereby fully unleashing its empowering effects on industrial chain resilience [2]. Compared with highly standardized industries such as manufacturing, the unique characteristics of fisheries production further reinforces this regional dependency. Fishing activities are often conducted in remote waters, which places extremely high demands on the accessibility of digital infrastructure. Meanwhile, the industry is characterized by long and vulnerable supply chains, while the intelligent operation of the entire process—from aquaculture to cold-chain logistics—relies heavily on system integration capabilities, which are precisely the strengths of the eastern regions with their complete industrial chains and well-developed supporting infrastructure.

7. Conclusions

Based on the above research conclusions, the following recommendations are put forward to better leverage the enhancing effect of AI on the resilience of the fishery industry chain and promote the high-quality development of the marine fishery economy:

First, promote the precise application of AI across the entire fisheries industrial chain. In the aquaculture sector, policymakers should support the diffusion of intelligent farming equipment based on machine vision and sensor networks to achieve automated management of water quality monitoring, feed delivery, and disease early warning. In the circulation sector, efforts should be directed toward establishing AI-enabled intelligent cold-chain logistics scheduling platforms for aquatic products, thereby optimizing inventory allocation and transportation routes while reducing product losses. In the sales sector, the development of smart traceability systems for fishery products should be encouraged to enhance brand value and market competitiveness. By embedding AI technologies into key nodes of the industrial chain, the overall resilience of the fisheries industrial chain can be systematically strengthened.

Second, implement regionally differentiated policies. Given their relatively advanced digital infrastructure, eastern coastal regions should prioritize the development of intelligent deep-sea fishing equipment, the construction of AI-driven fisheries big data platforms, and the extension of high-end segments within the industrial chain, thereby fostering regional hubs for fisheries technological innovation. In contrast, central, western, and less-developed fisheries regions should focus on addressing deficiencies in digital infrastructure, lowering the adoption threshold of AI technologies through fiscal subsidies and technical assistance, and increasing investment in science and technology.

Third, leverage institutional innovation to jointly unlock the resilience potential of industrial structure upgrading and optimal resource allocation efficiency. Along the industrial structure upgrading pathway, governments should provide support for advanced processing technology transformation and implement tax incentives to facilitate the transition of fisheries enterprises toward higher value-added segments, thereby strengthening value creation capabilities within the industrial chain. Along the resource allocation pathway, cross-entity data-sharing mechanisms and AI-driven dynamic scheduling platforms should be established to achieve the precise matching and efficient utilization of key production factors, including fishing quotas, aquaculture sea-use rights, and cold-chain storage resources.

Although this study has employed multiple tests to ensure the robustness of its conclusions, certain limitations remain.

First, the limited coverage of evaluation dimensions may induce estimation bias in resilience assessment. Industrial chain resilience is a multi-faceted concept that encompasses economic, ecological, and social attributes. This study primarily focuses on the economic dimension, which effectively reflects the growth and recovery capacity of fishery industrial chains. Nevertheless, it fails to fully capture resource sustainability and social adaptability, resulting in conclusions that are confined largely to the interpretation of economic resilience. To address this gap, future research may employ micro-level census data to construct multi-dimensional indicators at the fisherman and firm levels. By integrating economic, ecological, and social dimensions, a comprehensive resilience index can be developed to achieve a more holistic and accurate characterization of fishery industrial chain resilience.

Second, the depth of the transmission mechanism analysis can be further improved. This study investigates two mediating pathways through which AI shapes industrial chain resilience: industrial structure upgrading and resource allocation efficiency. The empirical findings confirm that AI strengthens industrial chain resilience by promoting structural optimization and resource allocation improvement. However, AI-enabled resilience enhancement is essentially a complex, multi-layered dynamic process. The current analytical framework does not incorporate potential mechanisms such as technology diffusion, factor restructuring, risk early warning, and collaborative governance, which limits the comprehensiveness and depth of our understanding of AI’s underlying influencing logic. Future studies may adopt firm-level patent data and input–output tables to quantitatively measure technology diffusion and factor restructuring effects. A multiple parallel mediation model can be constructed to systematically disentangle the diverse influencing pathways of AI empowerment and quantify their relative contributions.

Third, the exclusive focus on positive technological effects overlooks real-world adaptability constraints. This study emphasizes the beneficial impacts of AI on resilience enhancement. In practice, however, inadequate technological infrastructure, poor technology adaptability, and regional digital divides may substantially weaken or even offset technological dividends. This positive analytical perspective reduces the generalizability of the conclusions across regions with heterogeneous development conditions. Future research can adopt threshold regression and quantile regression models, employing digital infrastructure quality and technology adoption intensity as threshold variables to identify the critical conditions governing the shift in AI effects from positive to negative. Furthermore, combined with survey data and quasi-natural experiments in less developed fishery regions, subsequent studies can evaluate the heterogeneous impacts of digital divides on AI policy effectiveness. Such evidence can provide more nuanced empirical support for the formulation of differentiated industrial policies.