在人工智能背景下，深度伪造信息颠覆已有认知，引发信任危机，激化社会矛盾，精准地捕捉影响深度伪造信息分享意愿的关键要素和核心路径，能够有效地提高深度伪造信息治理的精度和效度。基于此，文章融合信息生态理论和SOR理论，采用问卷调查法收集了446份有效样本，运用多元线性回归分析探究单个变量与结果变量之间的因果关系，在此基础上，利用fsQCA方法探究多个变量对结果变量的组合效应，以期揭示深度伪造信息分享意愿的复杂因果驱动路径。研究发现：舆论环境、情感诉求、感知信息可信、信息权威性对深度伪造信息分享意愿具有正向影响，监管环境和信息危害性对深度伪造信息分享意愿具有负向影响。各变量均不构成高水平/低水平深度伪造信息分享意愿的必要条件，得到了9条驱动高水平深度伪造信息分享意愿的组态路径和10条驱动低水平深度伪造信息分享意愿的组态路径，据此提炼出3种高水平深度伪造信息分享意愿的驱动模式（低害驱动模式、权威诱导模式和媒体助推模式）和2种低水平深度伪造信息分享意愿的抑制模式（危害抑制模式和监管约束模式）。

In the context of artificial intelligence, deepfake information subverts existing cognition, triggers trust crises, and intensifies social conflicts. This article accurately captures the key elements and core paths that influence the willingness to share deepfake information. It can effectively improve the accuracy and effectiveness of deep forgery information governance. Based on this, integrating information ecology theory and SOR theory, this article collected 446 valid samples using a questionnaire survey method. It uses multiple linear regression analysis to explore the causal relationship between individual variables and outcome variables. On this basis, this article uses the fsQCA method to explore the combined effects of multiple variables on the outcome variables, in order to reveal the complex causal driving path of the willingness to share deepfake information. This study found that the public opinion environment, emotional appeals, perceived information credibility, and information authority have a positive impact on the willingness to share deepfake information, while the regulatory environment and information harmfulness have a negative impact on the willingness to share deepfake information. None of the variables constitute a necessary condition for high-level/low-level deepfake information sharing willingness. This article obtained 9 configuration paths that drive the willingness to share high-level deepfake information and 10 configuration paths that drive the willingness to share low-level deepfake information. Based on this, this article distills three high-level driving modes of willingness to share deepfake information (low-harm driving mode, authority-induced mode, and media-boosted mode) and two low-level inhibitory modes of willingness to share deepfake information (harm-inhibiting mode and regulatory-constraint mode).