In addition to exploring how people's expectations and beliefs about healthcare information and services affect their actual health outcomes, this study aims to empirically analyze whether there is a Pygmalion effect in healthcare platforms using machine learning and natural language processing. Regression modeling will be used to provide treatment recommendations for more common diseases. We gathered a 10-month panel dataset from a major Chinese online medical portal, containing information from 10,243 physicians. We discovered a strong linear correlation between users' expectations for their final level of recovery and satisfaction and their access to doctors, medical information, treatment alternatives, and healthcare experiences. People's choice of therapy for more complicated illnesses, like heart valve lesions and breast cancer, should lean more away from conventional information sources. Patients' expectations and treatment adherence are strongly connected with the expectations of their doctors, and treatment outcomes are also significantly influenced by the beliefs and expectations of the patients themselves. Using sentiment analysis and multiple robustness polls of user ratings on healthcare platforms, we demonstrate that the treatment choices made by users are distributed linearly across various complexity levels of diseases. As a result, this study highlights the real influence of patient and physician expectations and beliefs on healthcare outcomes, proves the presence of the Pygmalion effect on healthcare platforms, and explores it for particular diseases. This has real-world implications for raising patient happiness, enhancing medical service quality, and strengthening the doctor-patient bond.