Purpose/Significance The review texts of online mental health services are analyzed to identify patients’ preferences for online mental health services and put forward countermeasures and suggestions. Method/Process Taking Yidianling platform as an example, the content analysis method is used to form a medical preference index set. The latent Dirichlet allocation (LDA) model is used to determine the preferred topic. Logistic regression classification and sentiment analysis are used to identify topics of text comments and calculate online sentiment scores of each topic. Finally, the interpretative phenomenological analysis (IPA) method is used to identify the subject weight of patients’ preferences in choosing doctors and put forward suggestions for improvement. Result/Conclusion The formation of patients’ preferences in choosing doctors for online mental health services is divided into 3 dimensions and 9 sub-themes, which can provide important guidance and references for improving the quality and satisfaction of online mental health services.