Quantity or efficiency? The impact of AI adoption on firm innovation: Evidence from Chinese listed companies

The rapid advancement of artificial intelligence (AI) is profoundly reshaping the landscape of innovation. Using a panel of 2975 Chinese listed companies from 2013 to 2024, this paper applies text-mining techniques to quantify firms’ AI adoption and systematically examines its effects on both innovation quantity and innovation efficiency. The results show that AI adoption markedly increases innovation quantity but simultaneously constrains improvements in innovation efficiency, thus generating a pronounced “quantity-efficiency” paradox. This paradox remains robust across a range of sensitivity analyses and endogeneity adjustments. Mechanism tests reveal three channels through which AI adoption impedes innovation efficiency: reducing the time efficiency of R&D personnel, amplifying managerial expansion, and weakening the absorptive capacity of technical staff. Heterogeneity analyses further indicate that these negative effects are most pronounced in firms with higher human capital intensity, non-state ownership, and technology-intensive activities. Additional evidence shows that although AI adoption improves overall operational efficiency, it diminishes the individual productivity and compensation of R&D employees and exacerbates race-to-the-bottom competition among peer firms, thereby generating substantial negative externalities. By documenting the dual impact of AI adoption on corporate innovation and uncovering its underlying mechanisms—spanning human capital utilization, organizational restructuring, and inter-firm competition—this paper provides a more comprehensive and nuanced understanding of the complex interplay between AI development and innovation performance.