Enhancing patient knowledge and behaviour through digital health communication: a systematic review and random-effects meta-analysis of mobile, web-based, social media, telehealth, and AI-enabled interventions.

Background: Digital technologies are increasingly used to support health communication, yet their impact on knowledge, behaviour, and health literacy outcomes remains unclear. Previous reviews have often conflated improvements in knowledge with digital health literacy, without carefully distinguishing skills from outcomes. Objective: To systematically review and synthesize evidence on the effectiveness of digital health communication interventions-including mobile apps, web platforms, social media, telehealth/video tools, and emerging AI-enabled approaches-on knowledge, behaviour, clinical, and empowerment outcomes, while distinguishing these outcomes from digital health literacy when directly measured. Methods: We searched PubMed, Embase, Scopus, CINAHL, Web of Science, and Cochrane Library (January 2015 to January 2026). Eligible studies included randomized and non-randomized trials evaluating digital interventions designed to improve health knowledge, literacy, behaviours, or related outcomes. Only English-language, peer-reviewed studies were included; grey literature, conference abstracts, dissertations, preprints, and unpublished studies were not systematically searched. Random-effects meta-analysis was performed where outcomes were comparable, with subgroup and sensitivity analyses. For subgroup analyses, overlapping multi-component interventions were assigned once using a prespecified dominant patient-facing modality rule based on the main user interface and delivery channel; AI-enabled status was applied only when AI materially informed personalization or decision support. Certainty of evidence was graded using GRADE. Results: Thirty studies met inclusion criteria, and 12 contributed to at least one pooled meta-analysis. Continuous knowledge/literacy outcomes showed a small, statistically non-significant pooled effect (SMD 0.25; 95% CI: -0.04 to 0.54) with substantial heterogeneity. Dichotomous behavioural outcomes showed a statistically significant improvement (OR 3.57; 95% CI: 1.60 to 7.97), but this estimate was based on only two studies and should be interpreted cautiously. Mobile and web-based interventions provided the most frequent directionally favourable evidence, whereas AI-enabled, social media, and telehealth/video modalities were sparsely represented. Clinical and empowerment outcomes were too heterogeneous for pooling and showed mixed findings. GRADE certainty was low for knowledge/literacy, clinical, and empowerment outcomes and moderate only for selected behavioural outcomes. Conclusion: Digital health communication interventions may support patient knowledge and selected health behaviours; however, the strength of evidence remains limited by high heterogeneity, small numbers of pooled studies, overlapping intervention features, risk of bias, variable outcome measures, and exclusion of grey literature and non-English studies. Stronger conclusions about effectiveness, particularly for AI-enabled, social media, and telehealth/video approaches, require adequately powered trials using standardized literacy, behavioural, clinical, and empowerment outcomes. Systematic review registration: PROSPERO registration number: CRD42024625595.