Given that traditional satisfaction assessment methods have limitations such as being sensitive to noise and having inaccurate attribute extraction when dealing with large-scale, multi-source and heterogeneous comment data, this paper integrates large-scale group decision-making theory to propose a group consensus-based dynamic assessment method for online medical satisfaction. This study first defines a new distance measure based on single-valued neutrosophic cosine similarity, enabling flexible handling of uncertain information through adjustable parameters, and constructs a satisfaction quantification model for single-valued neutrosophic sets as a novel measurement tool for internet medical service quality evaluation. It then adopts a multi-label matching clustering method for precise attribute matching of review data, and employs the Sentence-BERT model to vectorize and reduce the dimensionality of review texts, enhancing the accuracy of effective information extraction. Furthermore, the research proposes a satisfaction update strategy based on dynamic weight adjustment and consensus integration, embedding it into a robust preference learning model to enable dynamic evaluation of online medical satisfaction. Finally, a case study using 730,408 chronic disease reviews from the WeDoctor platform validates the effectiveness of the proposed method. This research not only refines the satisfaction evaluation mechanism for smart healthcare service quality but also provides a reliable decision-making reference for service optimization within internet healthcare platforms, industry regulation, and patient choice of medical care.