How AI Adoption Reduces the corporate zombification risk—— Evidence from China

Persistent corporate zombification poses a significant structural challenge in transition and developing economies, distorting firm survival dynamics and undermining allocative efficiency. While prevailing literature attributes this phenomenon primarily to institutional distortions, it largely overlooks how technological upgrading can reshape firm survival incentives by endogenously enhancing corporate governance. This paper contends that the adoption of Artificial Intelligence (AI) serves not merely as a production technology but as a governance-enhancing mechanism. It systematically alters the equilibrium incentives between firms and external supporters by reducing the marginal cost of enforceable governance and increasing the credibility of verifiable signals. To formalize this logic, we embed a dynamic signaling game into a continuous-time optimal control framework, elucidating the theoretical pathways through which AI influences zombification risk via both cost and benefit channels.Empirically, we construct a firm-level AI adoption index by integrating textual analysis of annual reports with AI-related patent data for Chinese A-share listed firms from 2012 to 2023. Using panel models with high-dimensional fixed effects, instrumental variable estimation, and a multi-period difference-in-differences design, we establish a robust causal effect: AI adoption significantly reduces the probability of corporate zombification. Mechanism analyses confirm the dual channels: on the cost side, AI narrows investment-efficiency gaps and improves disclosure transparency; on the benefit side, it alleviates credit misallocation and enhances total factor productivity—though its effect on subsidy misallocation remains limited, highlighting institutional rigidities. Heterogeneity analyses further reveal that the de-zombification effect is more pronounced among non-state-owned firms, those in less-marketized regions, non-high-tech industries.