JMIR Publications

ABSTRACT

Background:

Artificial intelligence (AI)–enabled chatbots and related conversational systems can facilitate human–computer interaction through natural language, personalization, and automated support. In pediatric health promotion, these tools have potential to provide scalable and flexible approaches to support physical activity (PA) and related lifestyle behavior change within family contexts. However, evidence specific to family-based AI and chatbot-supported interventions for children and adolescents remains limited.

Objective:

This scoping review aimed to provide an up-to-date overview of how family-based AI and chatbot-supported interventions are designed, delivered, and evaluated for PA and related lifestyle behaviors among children, adolescents, and families, with attention to technology characteristics, family engagement, delivery platforms, outcomes, and research gaps.

Methods:

In accordance with PRISMA-ScR guidelines, seven databases (PubMed, Web of Science, APA PsycINFO, Academic Search Complete, CINAHL Ultimate, IEEE Xplore, and Scopus) were searched through February, 2026. Two reviewers independently conducted title/abstract and full-text screening in Rayyan AI. Eligible studies involved children, adolescents, or families; evaluated an AI-enabled chatbot, conversational agent, or related digital system; and assessed PA, obesity, weight management, sedentary behavior, or related lifestyle outcomes. Data were extracted and synthesized narratively in accordance with research objectives.

Results:

Of 2,730 records identified, 14 studies met inclusion criteria. Most were published in 2023 or later (12/14, 85.7%) and spanned 10 countries. Mobile app delivery was most common (9/14, 64.3%). AI approaches included rule-based chatbots, hybrid personalization systems, generative AI, and single studies using recommender systems or computer vision. 10 studies (71.4%) included a specific parent, caregiver, or family component. Common intervention features were personalized feedback (10/14, 71.4%), self-monitoring (9/14, 64.3%), and education (9/14, 64.3%). PA was the most common target behavior, often within broader obesity or healthy lifestyle interventions. Among eight studies reporting direct PA or fitness outcomes, four showed significant improvement, two found no significant change, and two reported mixed or indirect findings. Feasibility, acceptability, usability, and engagement findings were generally favorable across studies, but cultural tailoring was reported in only one study.

Conclusions:

Family-based AI and chatbot-supported interventions for pediatric PA and lifestyle behaviors represent a rapidly emerging but still early-stage field. Current evidence suggests promise for scalable and engaging support, but stronger trials, clearer reporting, culturally responsive designs, and more consistent family-centered evaluations are needed.