| 李秋静,朱晓梅,刘斌.生成式人工智能与数字产品出口——来自生成式人工智能服务备案的证据[J].数量经济技术经济研究,2026,(7):133-158 | | 生成式人工智能与数字产品出口——来自生成式人工智能服务备案的证据 | | Generative AI and Digital Product Exports: Evidence from Generative AI Service Registration Data | | | | DOI： | | 中文关键词: 生成式人工智能 数字产品出口 数字技术 | | 英文关键词: Generative AI Digital Product Export Digital Technology | | 基金项目: | | | 中文摘要: | | 作为新一轮科技革命的核心驱动力量，生成式人工智能正深刻重塑国际贸易格局。本文使用生成式人工智能服务备案数据、地区层面出口数据和标准普尔Panjiva数据库提供的企业提单数据，并结合BERT-wwm-ext中文模型，在宏观地区和微观企业两个层面检验了生成式人工智能对数字产品出口的影响。研究发现，生成式人工智能显著促进了中国数字产品出口。机制分析表明，生成式人工智能通过三条渠道发挥作用：一是基于交互式能力优化跨境搜寻与匹配效率；二是基于生成式能力扩展出口产品种类；三是基于“通用式”能力提升出口产品质量。异质性分析表明，在技术层面，应用于电商、产品设计等商业场景以及提供网页访问方式的生成式人工智能，对数字产品出口的促进效应更为明显。在产品层面，生成式人工智能对非标准化和交互型数字产品的出口促进效应更大。在地区层面，在电子商务发展水平较高、算力水平较高的地区以及人工智能试验区，生成式人工智能促进数字产品出口的效应更显著。本文为理解生成式人工智能对国际贸易的影响提供了经验证据，并为中国推动生成式人工智能发展提供了参考。 | | 英文摘要: | | As a core driving force of the new wave of technological revolution, generative artificial intelligence (AI) is profoundly reshaping the global trade landscape. Unlike previous generations of AI technologies, generative AI features stronger interactive capabilities, broader general-purpose applicability, and more flexible content-generation functions, enabling it to exert deeper and more direct impacts on digital trade activities. Against this background, the exploration of whether and how generative AI affects the export of digital product is of considerable theoretical and practical significance to understand the trade effects of emerging digital technologies and promotethe high-quality development of China’s digital trade.
This paper empirically examines the impact of generative AI on China’s digital product exports at both the macro-regional and micro-firm levels. Specifically, we construct a novel measure of regional generative AI application intensity, based on the filing and registration data of generative AI services officially released by the Cyberspace Administration of China. To accurately identify digital products, we improve existing classification approaches by applying the BERT-wwm-ext Chinese language model to product text recognition. Combining this identification strategy with regional product-level export data from 2021 to 2025 and with more than one billion firm-product-level bill-of-lading records from the S&P Panjiva database, we establish a comprehensive empirical framework to investigate the export effects of generative AI.
The empirical results show that generative AI significantly promotes China’ s export of digital products. This conclusion remains robust even after a series of robustness checks, including alternative variable measurements, endogeneity corrections, and sample adjustments. The mechanism analysis reveals that generative AI affects the export of digital product through three major channels. First, generative AI improves the efficiency of cross-border search and matching thanks to its interactive capabilities. By enhancing firms’ ability to communicate with overseas clients, conduct multilingual interactions, and identify market demand more efficiently, generative AI reduces information asymmetry and lowers the transaction costs in international digital trade. Second, generative AI expands the variety of exported digital products, thanks to its generative capabilities. By accelerating product design, creative development, and content production, generative AI enables firms to diversify product offer and rapidly respond to the heterogeneity of global demand, thereby broadening the scope of digital product exports. Third, generative AI improves the quality of the exporteddigital products,thanks to its general-purpose capabilities. As a foundational technology that can be applied across multiple production and operational scenarios, generative AI enhances product optimization, service customization, and innovation efficiency, ultimately raising the quality competitiveness of exported digital products. The results of further heterogeneity analysis suggest that the export-promoting effect of generative AI is not homogeneous. At the technological level, compared to other scenarios or local deployment, the use of generative AI in commercial scenarios, such as e-commerce and product design, as well as the accessibility of AI services through web-based platforms, exhibit significantly stronger export-enhancing effects. At the product level, generative AI has a more pronounced impact on the export of non-standardized and interactive digital products, where customization and user interaction play a more central role. At the regional level, this positive effect is significantly stronger in those regions with higher levels of e-commerce development, stronger computing infrastructure, and established AI pilot zones.
This study makes three main contributions to the literature. First, by focusing on the technological characteristics of generative AI, it expands the theoretical understanding of how technological progress affects international trade, broadening the explanatory boundary of digital technology-driven export growth. Second, it systematically identifies the mechanisms through which generative AI affects the export of digital products, clarifying the roles of interactive capability, generative capability, and general-purpose capability. Third, it improves measurement strategies in two aspects: by enhancing the accuracy of digital product identification through the BERT-wwm-ext model, and by constructing a novel indicator of regional generative AI application intensity using official filing data, thus providing a methodological reference for future research.
This paper provides new empirical evidence on the trade effects of generative AI, offering important insights to understand the evolving relationship between frontier digital technologies and international trade. Moreover, it offers policy implications to promote the deep integration of generative AI into digital trade and enhance China’s international competitiveness in the digital economy. | | 查看全文 相关附件： 下载数据代码附录 |
| | |
