Due to the family-oriented cultural and legal context in China, understanding the difference between patients' and family members' experiences of psychiatric services not only enriches perspectives of service quality assessment, but also promotes service utilization. This study aimed to compare experiences of psychiatric inpatient services between patients and their family members in China. The study included 126 dyads of patients and family members consecutively recruited from the psychiatric inpatient department in a large hospital in China. The responsiveness performance questionnaire was used to measure the experiences of psychiatric inpatient services after patient discharge. After adjusting reporting heterogeneity based on vignettes, dyad difference was examined by intraclass correlation coefficients (ICCs) and paired Wilcoxon signed-rank tests with Bonferroni correction in multigroup testing. Subgroup analyses were conducted within strata of four selected clinical and socio-demographic factors, to test their influence on difference pattern of experiences. Poor consistency was found for all responsiveness items and the total scores among the 126 dyads and in most subgroup analyses (ICC < 0.6). Paired Wilcoxon signed-rank tests found that patients rated lower than their family members on the item of "asking user's opinions" in 126-dyad comparison (P < 0.05) and 3 subgroups related to severe mental disorders and income inequality after Bonferroni correction. Results reveal inconsistent experiences of psychiatric inpatient services within families in China. Moreover, when making medical decisions, family members' opinions, rather than patients', are more frequently taken into consideration, especially on conditions where imbalanced decision-making power exists between patients and their family members. In the future, user experience improvement should pay equal attention to patients and family members, and the benefits of family involvement and patients' rights of shared decision-making should be carefully balanced.