Interpersonal Neural Synchrony (INS) refers to the dynamic covariation of neural activity between interacting individuals, and serves as the core neural substrate for effective spoken communication. Despite the important foundation laid by classic theoretical frameworks (shared representation theory, predictive coding theory), current research still faces three unresolved core questions: the dynamic evolution mechanism of INS across communication stages, the synergistic mechanism between hierarchical linguistic information and contextual factors, and the association between the core cognitive mechanisms and neural substrates of INS. This review systematically synthesizes empirical evidence on the synchronization of multi-level linguistic representations (phonetic/prosodic, syntactic, semantic) and mentalizing-related language-context integration in spoken communication. Based on the Interactive Alignment Model and the Integrative Model of Pragmatic Competence, this paper proposes the Hierarchical-Dynamic Integration Model of INS in spoken communication, which elaborates on the cognitive-neural mechanisms underlying INS formation—jointly driven by two interdependent processes (hierarchical linguistic synchronization and dynamic language-context integration), regulated by three core cognitive processes (imitation, adaptation, and prediction), and realized relying on the coordinated activity of the left hemisphere classical language network, right hemisphere homologous regions, and Theory of Mind (ToM) network. This paper further analyzes the limitations of the model (ignoring multimodal non-linguistic information, language-attention interaction, and language-emotion synergy), and puts forward targeted future research directions, including model improvement, deepening of basic mechanisms, exploration of special phenomena, analysis of group characteristics, and interdisciplinary applications. This review enriches the theoretical system of the neural basis of human social communication, provides key theoretical support and neural constraints for cognitive computational modeling of spoken communication, and promotes interdisciplinary integration in the fields of neuroscience, human-computer interaction, and clinical communication intervention.