AI-assisted drawing therapy for children with ADHD: A randomised controlled trial

OBJECTIVE: This randomized controlled trial evaluated the efficacy of AI-assisted drawing therapy for children with ADHD, comparing its impact on core symptoms and functional impairments against traditional drawing interventions while assessing engagement and the rapeuticoutcomes. METHODS: Forty-one children (ages 7-10) with ADHD were randomized into AI-assisted (n = 19) or traditional drawing therapy (n = 22) groups (Trial registered in ChiCTR: ChiCTR2400089640). The AI intervention employed a process that captured children's drawings, converted them into keywords, generated AI responses based on those keywords, and provided feedback to the children, while the control group received therapist-guided verbal feedback. SNAP-IV-26(Swanson, Nolan and Pelham-IV rating scales) and WFIRS-P(Weiss Functional Impairment Rating Scales-Parent) scales assessed at baseline (T1) and post-intervention (T2) at a 24-week interval, with data analyzed using independent t-tests (p < 0.05). RESULTS: Both AI-assisted and traditional drawing therapies significantly reduced core ADHD symptoms. In the before-and-after comparison within the group, the AI group showed targeted improvements in attention (p = 0.039) and oppositional defiant behaviour (p = 0.025), while traditional therapy produced broader symptom reductions across all domains (p < 0.05). Between-group comparisons showed that the AI-assisted group achieved greater improvement in inattention (Δ = -1.09, p = 0.035), whereas the traditional drawing group improved more on oppositional defiant behavior (Δ=+1.706, p = 0.010) and the SNAP-IV total score (Δ=+4.777, p = 0.041). On the WFIRS-P, the AI group showed advantages for the Family domain (p = 0.041) and the total score (p = 0.047), while Self-Concept favored the traditional drawing group (p = 0.036). CONCLUSION: AI-assisted, drawing-based art therapy produced greater improvement in the pre-specified primary outcome (SNAP-IV inattention) and select WFIRS-P domains (e.g., Family, total score), whereas conventional drawing therapy yielded broader reductions in oppositional defiant symptoms and the overall SNAP-IV severity. Effects were modest and domain-specific; larger, multi-center trials are needed to confirm efficacy and identify which children benefit most.