Based on context-driven innovation theory and task-technology fit (TTF) theory, this study focuses on knowledge-intensive enterprises and adopts a mixed-method research design to explore the mechanism through which the fit among generative AI technology, domain knowledge, and application contexts influences enterprise innovation performance. Through multiple case studies, a “technology-knowledge-context-innovation” theoretical framework for generative AI is constructed, which is further empirically tested using structural equation modeling (SEM). The findings indicate that different enterprises present distinct “technology-knowledge-context” fit paradigms. Furthermore, generative AI technology characteristics, domain knowledge characteristics, and application context characteristics all exert significant positive effects on “technology-knowledge-context” fit. In addition, “technology-knowledge-context” fit not only directly enhances enterprise innovation performance, but also indirectly promotes innovation performance by increasing the application intensity of generative AI.