As AI for Science (AI4S) emerges as a new frontier in global technological competition, clarifying the micromechanisms of how Artificial Intelligence (AI) reshapes the R&D innovation paradigm has become a critical issue This paper aims to reveal the dynamic effects and complex mechanisms of AIdriven R&D innovation This study constructs a “TechnologyTaskAbility” (TTA) analytical framework to argue that AI profoundly reshapes the R&D innovation ecosystem by reorganizing, rather than simply accelerating, R&D tasks Based on patent data from the global carbon nanotube field, this paper employs a combination of methods, including technology cooccurrence network analysis, the timevarying parameter vector autoregression model, impulse response analysis, and wavelet coherence analysis, to empirically examine the dynamic evolution of the technology network before and after the integration of AI The findings indicate that: first, the integration of AI significantly enhances the overall connectivity of the technology network, with AI itself becoming a key technology spillover source and playing the role of a “leader” in the innovation network Second, by automating tasks such as materials characterization, AI causes traditional analytical technologies, such as materials testing and analysis, to shift from a “leader” to a “follower” role within the network, providing direct empirical evidence for the task reorganization mechanism proposed by the TTA framework Third, impulse response and wavelet coherence analyses further confirm that AI has a significant and dynamic positive driving effect on the development of carbon nanotube technology Furthermore, by analyzing the interactive logic between AI and new materials R&D innovation, the paper proposes principles of “stratified governance” and “segmented governance” for the R&D innovation ecosystem