Child and Adolescent Mental Health Misinformation on TikTok: Cross-Sectional Mixed Methods Study.

BACKGROUND: TikTok is an important source of mental health information for children, adolescents, and young adults, but the accuracy of psychoeducational content remains uncertain. Misleading information may be particularly consequential for younger users, who may have limited digital and health literacy. OBJECTIVE: This study aimed to characterize misinformation and disinformation in TikTok content related to child and adolescent mental health, identify mental health topics disproportionately represented in misleading content, examine whether platform-observable video characteristics could discriminate videos containing misinformation, and explore how viewers responded to this content. METHODS: We conducted a cross-sectional mixed methods secondary analysis of 122 child and adolescent mental health videos identified from a previously collected dataset of 1000 publicly available TikTok videos. Videos were independently coded for content, intent, authenticity, clinical topic, presenter characteristics, and engagement. An L2-penalized logistic regression using only platform-observable characteristics was evaluated using stratified 5-fold cross-validation; 95% CIs for predictive metrics were estimated using 5000 stratified bootstrap resamples of out-of-fold predictions. Disinformation was analyzed descriptively because only 2 cases were identified. An exploratory thematic analysis was conducted on 116 selected viewer comments. RESULTS: Misinformation was identified in 21 of 122 (17.2%; 95% CI 11.5%-24.9%) videos, whereas disinformation occurred in 2 of 122 (1.6%; 95% CI 0.5%-5.8%). Misinformation was most frequent among videos addressing autism spectrum disorders (11/27, 40.7%), impulse control disorders (1/3, 33.3%), and mood disorders (4/13, 30.8%). Professional credentials were not disclosed in 63 of 122 (51.6%) videos, and only 9 of 122 (7.4%) referenced an external scientific source. The observable-feature model demonstrated modest discrimination, with an area under the receiver operating characteristic curve of 0.70 (95% CI 0.56-0.82), balanced accuracy of 0.69 (95% CI 0.58-0.79), recall of 0.67 (95% CI 0.48-0.86), and F1-score of 0.44 (95% CI 0.32-0.56). Qualitative analysis identified 5 recurring patterns in viewer responses: sharing and validation of lived experiences, discussion of diagnosis and misdiagnosis, frustration with mental health care, peer support and encouragement, and exchanges about treatment and medication experiences. CONCLUSIONS: A substantial minority of the analyzed TikTok videos contained misinformation, with particularly high proportions among some neurodevelopmental and mood-related topics. Platform-observable characteristics provided only modest discrimination of misinformation, underscoring the difficulty of identifying misleading mental health content from surface-level features alone. Viewer comments further showed that TikTok mental health content is embedded within broader processes of self-interpretation, diagnostic discussion, peer validation, and treatment-related exchange. These findings support the need for developmentally appropriate digital health literacy initiatives and accurate, transparently sourced psychoeducational content for young social media users.