Aug 2026· Frontiers in Psychiatry· 0 citations· 52 references
TL;DR
Based on the findings, AI seems highly promising for patient-specific ASD therapy via proactive, data-driven scaffolding, but more RCTs and crucial augmentation of representation gaps concerning adult and female ASD phenotype studies are required.
Abstract
The increased prevalence of ASD has generated a pressing demand for flexible therapeutic and educational tools. AI has been suggested as a potential bridge to this gap, but the translation from a model to a clinical application necessitates rigorous assessment. The purpose of the present review is to compile and examine the existing literature to demonstrate AI interventions that have advanced from a proposed model to being used with human participants.
We systematically searched five databases (Embase, PubMed, ScienceDirect, IEEE Xplore, Web of Science; Jan 2016–Dec 2025) for AI-based ASD interventions. Two reviewers independently assessed eligibility. Inclusion criteria were as followed: (1) participants with confirmed ASD diagnoses; (2) an intervention sample size of N ≥ 6; (3) AI as a central therapeutic, educational, or rehabilitative component; and (4) multi-session protocols with specified timeframes. Study types ranged from system development and feasibility trials to RCTs.
14 studies met inclusion criteria. AI (e.g. robotics, VR, and wearables) functioned as a social mediator, improving social-emotional outcomes (e.g., ADOS, SRS scores) by reducing cognitive load. Significant mechanisms included real-time task adaptation and precise behavioral monitoring via e.g. eye-tracking. However, significant heterogeneity was observed in intervention dosage (median 4–12 hours). Most studies were limited by small, male-dominated samples (N < 20) and a total absence of adult participants.
Based on the findings, AI seems highly promising for patient-specific ASD therapy via proactive, data-driven scaffolding. In order to move toward implementation of AI in ASD care, more RCTs and crucial augmentation of representation gaps concerning adult and female ASD phenotype studies are required.
Artificial intelligence (AI) has been increasingly integrated into autism interventions to support personalization and scalability; however, the strength of empirical evidence supporting these approaches remains unclear. In this systematic review, we synthesized and critically appraised experimental studies evaluating AI-based interventions for autistic individuals with a specific focus on intervention outcomes and methodological rigor. Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines and a preregistered PROSPERO protocol, we searched databases and identified 13 eligible studies, including randomized controlled trials, quasi-experimental group designs, and single-case experimental designs. Reviewed studies targeted social engagement, communication, emotion recognition, empathy, adaptive participation, and symbolic play and predominantly employed human-in-the-loop models involving parents, educators, or clinicians. Methodological quality was evaluated using design-appropriate quality indicators and What Works Clearinghouse (WWC) standards. Although most studies reported positive short-term effects on participant-level outcomes, rigor varied considerably. Only two studies met WWC standards, five met standards with reservations, and six did not meet standards due to limitations related to experimental control, fidelity reporting, outcome measurement, or data adequacy. Overall, AI-based interventions show promise as tools to augment human-delivered autism interventions, but the current evidence base is preliminary. More rigorous research is needed to establish effectiveness and inform implementation in autism services. Lay Abstract Artificial intelligence, often called AI, is increasingly being used to support services for autistic children. AI tools can help adults such as parents, teachers, and therapists personalize instruction, track progress, and provide feedback during everyday activities. However, it is not yet clear how strong the research evidence is for AI-based interventions. In this review, we examined studies that tested AI-supported interventions designed to improve learning and behavior outcomes for autistic individuals. We carefully reviewed 13 studies and evaluated how well these studies were designed and conducted. The studies focused on areas such as social engagement, communication, emotion understanding, empathy, and participation in daily routines. Most interventions used AI to support, rather than replace, adult guidance. Although many studies reported positive short-term improvements, we found that many had important limitations in their research design. Only a small number of studies met strong standards for research quality. This means that more careful and well-designed studies are needed before AI-based interventions can be widely recommended. Overall, AI shows promise as a tool to support autism intervention, but stronger evidence is needed to understand when, how, and for whom these tools are most helpful.
Yusuf Akemoğlu, Emily Brooke Roberts, Jinjun Xiong· Autism· 0 citations
The results suggest that the use of AI to detect ASD is not yet ready for clinical use, and transparent reporting of validation, independent external cohorts, diverse datasets, and prospective testing of positive predictive value in realistic population prevalence should be the focus of future research.
Shivani Pant, A. Gehlot, Neha Singh et al.· International Journal for Gl...· 0 citations
Artificial intelligence (AI)-driven technologies are increasingly deployed in pediatric rehabilitation for children with neurodevelopmental and motor disorders, including attention-deficit/hyperactivity disorder (ADHD), autism spectrum disorder (ASD), and cerebral palsy (CP). Despite the rapid increase in AI-enabled interventions across clinical and home settings, the evidence base for clinical applicability, effectiveness, and implementation requirements remains insufficiently characterized. This systematic review and meta-analysis aimed to: (i) describe the types of AI technologies and intervention models used across diagnostic groups and settings; (ii) summarize reported effects on child outcomes; and (iii) critically appraise clinical applicability, including feasibility, acceptability, safety, equity considerations, and key implementation requirements. A systematic review and meta-analysis of randomized controlled trials (RCTs) evaluating AI-driven or AI-assisted interventions in children and adolescents (≤18 years of age) with ADHD, ASD, or CP was conducted. Five databases (MEDLINE, Embase, PsycINFO, CINAHL, Web of Science) were searched for studies published between 2020 and 2025. Two independent reviewers screened the studies, extracted data, and assessed the risk of bias using the Cochrane Risk of Bias 2 (RoB2) tool. Effect sizes were calculated as Hedges’
g
using random-effects models, with subgroup analyses conducted for three intervention–condition pairs. Fifteen RCTs were included, enrolling approximately 1039 participants (ages 3-18 years) across nine countries. Six AI technology categories were identified: adaptive digital therapeutics (
n
= 3), robot-assisted therapy (
n
= 6), virtual reality systems (
n
= 2), robotic exoskeletons (
n
= 2), neurofeedback (
n
= 1), and mobile cognitive training (
n
= 1). The International Classification of Functioning, Disability and Health for Children and Youth (ICF-CY) mapping revealed that 62.5% of studies targeted body functions outcomes and 37.5% targeted activities outcomes. A meta-analysis of eight studies yielded a statistically significant overall pooled effect [Hedges’
g
= 0.26 (95% confidence interval: 0.04, 0.49),
P
= 0.02,
I
2
= 40%]. Clinical applicability varied substantially: digital therapeutics demonstrated highest feasibility and scalability for home-based delivery, while robotic exoskeletons showed limited accessibility due to cost and infrastructure requirements. Equity concerns were identified, with 14 of 15 studies conducted in high-income countries. AI-driven interventions produce statistically significant, small-to-moderate benefits across ADHD, ASD, and CP, with effect sizes varying by technology type and diagnostic group. These technologies should be considered adjunctive tools within multimodal rehabilitation programs rather than standalone replacements for conventional therapies. Future research priorities include larger trials with long-term follow-up, standardized functional outcome measures, equity-focused study designs in diverse populations and settings, and implementation research examining cost-effectiveness and integration into existing service systems.
M. Alghadier, Rafif Alsedrani, Noor Alhabib et al.· Journal of Disability Resear...· 0 citations
ABSTRACT Background Although speech and language therapy (SLT) is central to post‐stroke aphasia rehabilitation, global SLT provision often falls short of recommended dosages. In response, interest has grown in technology‐based interventions, including therapy software, virtual reality (VR) and artificial intelligence (AI) tools. However, current evidence for the effectiveness of technology is fragmented, and no recent review offers a comprehensive synthesis across these three modalities. Aim This review examined the range of technologies used in aphasia assessment and therapy and summarised their effectiveness across different intervention targets. The two research questions were: (1) What types of technology have been investigated for assessing and treating people with aphasia (PWA)? (2) How effective is the use of technology in the assessment and treatment of PWA? Methods A systematic search of four databases (PubMed, PsycINFO, Web of Science and Scopus) covering the period 2013 to May 2026 identified 67 included studies, of which 14 were randomised controlled trials. Studies reporting quantitative outcomes, were peer‐reviewed, and focused on technology‐based intervention for PWA were eligible. Quality was appraised using the NICE checklist. The GRADE framework was applied to evaluate certainty of evidence for each intervention target. Findings were then synthesised narratively due to heterogeneity across study designs, and outcome measures. Results Three technology types were identified: computerised speech and language therapy (CSLT) (38 studies), VR (17 studies) and AI (13 studies). AI was used predominantly for aphasia assessment and classification. The strongest and most consistent evidence related to word‐finding, where high certainty of evidence was supported by multiple RCTs delivering therapy at or above the recommended 20‐h threshold. For language production and comprehension, functional communication, and reading, outcomes were more variable, reflecting moderate certainty of evidence, and inconsistent dose adherence. Writing interventions received a low certainty rating, reflecting small samples, limited blinding and task‐specific rather than generalised gains. Across domains, higher‐dose studies were consistently associated with better outcomes, which may suggest that technology functions primarily as a tool to enable high‐intensity practice rather than as an independently effective treatment ingredient. Conclusion CSLT, VR and AI tools show promise as adjuncts to face‐to‐face SLT for aphasia assessment and rehabilitation. Word‐finding interventions delivered at recommended doses have the strongest evidence base. Some studies did not use technology to support the recommended therapy dose. For other intervention targets, larger, higher‐dose trials are needed. Future research should also examine whether integrating different technology types could offer additional clinical benefit. WHAT THIS PAPER ADDS What is already known about the subject Previous systematic reviews have demonstrated the emerging role of technology in aphasia rehabilitation, with earlier work focusing primarily on computer‐based therapy or AI technologies. However, these reviews were either narrow in scope (targeted specific technology type), or outdated. What this study adds to the existing knowledge This review provides an updated, cross-technology synthesis encompassing AI, virtual reality, and computerised speech‐and‐language therapy. It outlines how these tools were applied within the studies in the literature. The review also identifies persistent limitations in therapy dosage across studies, underscoring the need for future higher‐dose trials to confirm the certainty of evidence across different intervention targets. What are the clinical implications of this study? The growing evidence for computerised speech and language therapy, virtual reality, and artificial intelligence tools continues to support their role as adjuncts to face to face SLT, particularly for language assessment and targeted word‐finding interventions. These technologies may extend therapy provision beyond clinical hours, enable therapeutic doses of practice to be achieved, and improve consistency in assessment procedures.
Bader Alhejji, S. Cunningham, Samantha Dorney et al.· International journal of lan...· 0 citations
Artificial intelligence (AI) is transforming mental health care by enabling novel approaches to assessment, outcome prediction, and treatment. However, the rapid growth of AI tools has outpaced evidence synthesis on their realworld clinical value, implementation barriers, and ethical risks, leaving clinicians and policymakers without clear guidance for responsible integration. This narrative review aimed to examine current AI applications in mental health care, identify their clinical benefits, implementation barriers, and ethical challenges, and define conditions for responsible integration that preserve the patient-clinician relationship. A structured literature search of PubMed, PsycINFO, and Scopus was conducted for peer reviewed English language articles published between January 2019 and March 2026. Studies addressing AI applications, benefits, implementation, or ethics in mental health were included and synthesised narratively into thematic categories. The review found that AI tools—for example, selfreferral chatbots that reduced waiting times and increased treatment uptake—provide expanded 24/7 access, improved clinical efficiency, and potential for individualised personalisation. Predictive models showed promise for treatment selection and risk stratification, while natural language processing unlocked unstructured clinical data. However, concerns included patient data privacy, algorithmic bias that may worsen existing inequities, potential erosion of the therapeutic relationship, and mixed acceptance among clinicians and patients. Many AI applications remain experimental, and regulatory frameworks have not kept pace with technological developments. This review did not include formal quality appraisal or quantitative synthesis; the evidence is limited by short followup periods, high dropout in chatbot trials, and a predominance of studies from highincome countries. Future research should employ longitudinal, codesigned, mixedmethods designs and pragmatic trials that evaluate clinical outcomes alongside equity, user trust, and the preservation of empathic, humanled care—rather than relying solely on uncontrolled implementation studies. AI should be responsibly integrated to augment, not replace, the clinical workforce. Successful application requires a balance between technological advancement, patient protection, and preservation of the patient-clinician relationship.
Shizal Nawaz, Laiba Nawaz, Hasnain Ali et al.· Digital Medicine· 0 citations
ABSTRACT Objective: to map the scientific evidence on health interventions for children with Autism Spectrum Disorder (ASD) during hospitalization. Method: this is a scoping review developed in accordance with the recommendations of the Joanna Briggs Institute, conducted in February 2025 across the Scopus, PubMed/MEDLINE, Web of Science, LILACS, and BDENF databases. Included were primary studies, reviews, theses, dissertations, and experience reports addressing autism and care strategies in hospital settings, available in full, published in any language, and with no time restriction. Results: 12 studies comprised the final sample. Following analysis, the identified interventions were grouped into four categories: 1) care strategies in inpatient psychiatric settings; 2) care strategies in general hospital settings; 3) care in the emergency department; 4) care in the surgical department. Conclusion: the study identified intervention strategies that can improve the hospitalization experience of children with ASD. These interventions show potential to significantly enhance the hospital experience of children with ASD and the effectiveness of their treatment.
Benedita Shirley Carlos Rosa, Angelina Germana Jones, Francisco Mardones dos Santos Bernardo et al.· Texto & Contexto - Enfer...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
