Supporting Social and Communication Outcomes in Children with Autism Spectrum Disorder via Mobile Device-Based Interventions: A Systematic Review

Abstract

Children diagnosed with Autism Spectrum Disorder (ASD) often encounter significant challenges in social interaction and communication. The integration of portable devices (e.g., mobile phones and tablets) into educational frameworks is transforming the support landscape for students with ASD, with these devices serving as feasible and potentially advantageous delivery platforms for interventions targeting selected, directly trained outcomes. The objectives of this review are threefold: (a) to synthesize peer-reviewed research regarding the use of portable devices to enhance social skills in children with ASD; (b) to evaluate and analyze the efficacy of these digital interventions; (c) to identify current limitations in the literature while providing evidence-based recommendations for practitioners, educators, and future research. The review was conducted and reported in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. A systematic search was conducted using the following electronic databases: (a) PsycINFO, (b) Education Resources Information Center (ERIC), (c) Scopus, and (d) Web of Science. Fourteen studies were included, comprising a total of 201 children: 154 children with ASD and 47 typically developing children who served as an active control group in one study. Ten studies employed single-case experimental designs, whereas four used group-based research designs. Given the heterogeneity in study designs and outcome measures, the findings were synthesized narratively rather than through a meta-analysis. Data on intervention components, implementation strategies, and outcomes were extracted. The findings indicate that portable devices may offer a feasible and potentially advantageous means of delivering interventions for selected, directly trained outcomes in children with ASD; however, the strength and consistency of the evidence varied across studies and research designs. Several studies reported improvements in targeted communication, social–emotional, and interpersonal outcomes, whereas evidence regarding the generalization and maintenance of these improvements remained comparatively limited. The current evidence base does not support firm conclusions regarding their overall effectiveness in improving social skills. The available literature also exhibits considerable variability in sample sizes, study designs, outcome measures, and follow-up durations. Future research should prioritize rigorous designs, including studies with larger samples and long-term follow-up to assess the sustainability of outcomes.

1. Introduction

According to the Centers for Disease Control and Prevention (CDC), the national public health agency of the United States, approximately 1 in 31 (3.2%) eight-year-old children have been diagnosed with Autism Spectrum Disorder (ASD) (Shaw et al., 2025), with boys being approximately four times more likely to be affected than girls (Abdul Hakim et al., 2021; Berenguer et al., 2020; Miningrum et al., 2021). Children with ASD commonly experience significant difficulties in social interaction, largely attributable to deficits in social functioning (American Psychiatric Association, 2013).

The construct of social skills encompasses a set of learned behaviors that enable individuals to engage effectively in social interactions. It reflects a dynamic process involving the individual, the communicative intent, and the surrounding social context. More specifically, social skills comprise behaviors that facilitate adaptive interpersonal engagement, as well as the regulation and avoidance of socially inappropriate responses (Moosapoor, 2023). Furthermore, these skills involve the capacity to navigate situational demands that elicit and sustain social interaction, thereby requiring a degree of social competence. Intervention outcomes in this field span a range of developmental domains, reflecting the multifaceted nature of children’s social, communicative, and cognitive functioning. For the purposes of the present review, the social skill outcomes identified across the included studies were subsequently grouped into three broad domains to facilitate their organization and synthesis. The first domain, communication and pragmatic language, encompasses the functional use of language in social contexts, including requesting, question answering and conversational responding, as well as prosody. The second domain, social–emotional and interpersonal skills, includes abilities related to understanding and engaging with others, such as emotion recognition, understanding social norms, and social engagement. The third domain, cognitive and pre-academic skills, includes foundational abilities that support learning and communication, such as receptive vocabulary, defined as the comprehension of spoken words and language. This taxonomy provides a structured framework for examining intervention outcomes across studies.

A central barrier to the development of social skills in children with ASD lies in the communication difficulties that permeate their everyday experiences. These individuals are significantly less likely to initiate social interactions (Pavlou & Garmpis, 2025), a pattern largely driven by impairments in joint attention and peer engagement, which constitute foundational mechanisms of early social reciprocity. Such disruptions may compromise social orienting, resulting in reduced responsiveness to social cues and the impression of diminished awareness of others. Consequently, limitations in initiating and sustaining reciprocal play emerge, restricting opportunities for meaningful peer engagement and thereby impeding the formation and maintenance of stable social relationships (Kasari & Patterson, 2012). In addition, communication in ASD is often characterized by atypical prosodic features, including variations in intensity, rhythm, speech rate, and intonation patterns (Qualls & Corbett, 2017). Within this context, social routines assume a critical role. Such routines consist of structured, mutually engaging activities between two or more individuals, during which the child’s attention is directed toward the social partner or the interaction itself rather than toward objects (Rogers & Dawson, 2010). When these routines are experienced as intrinsically rewarding, they can provide meaningful context for the acquisition and generalization of social communication skills (Ramseur et al., 2019).

Another domain that substantially influences everyday social functioning is that of social norms. These norms constitute implicit regulatory frameworks that govern behavior within a social group. They play a pivotal role in shaping actions, cognitive processes, and interpersonal interactions by establishing shared expectations regarding acceptable and non-compliant behavior. Understanding the mechanisms through which such norms emerge and exert influence is essential for a comprehensive account of social functioning (Costescu et al., 2025). Children with ASD are less likely to conform to social norms compared to their typically developing peers, which may hinder their ability to interpret social cues and pose challenges in the formation and maintenance of peer relationships (Yafai et al., 2014).

Language development represents another critical factor influencing communication outcomes in children with ASD (A. S. Chan et al., 2005). Early interventions targeting language acquisition—particularly during the preschool years—have been shown to significantly enhance linguistic competencies. Indeed, the implementation of early behavioral interventions is associated with more favorable developmental trajectories, including improvements in social adaptation, communication skills, and subsequent language development (Hesami et al., 2025). In the broader preschool population, speech and language development has also been positively associated with school readiness, underscoring the developmental importance of early language skills (Koçak & Korkmaz, 2023).

Importantly, the development of language skills alone is insufficient; their functional use is equally essential. Functional language encompasses the ability to express needs through requests, respond appropriately to questions, recognize opportunities for turn-taking, and engage in increasingly complex verbal behaviors. These competencies collectively contribute to both social and cognitive development (Ganz et al., 2017).

Within the domains of linguistics and speech–language pathology, the term receptive lexicon refers to the set of words an individual can comprehend through listening or reading, even in the absence of active production. Children with ASD often exhibit difficulties in acquiring new vocabulary (Pickles et al., 2014), which may constrain lexical development and, in turn, negatively affect social functioning (El Shemy et al., 2025). It is estimated that approximately 25% to 30% of children with autism do not develop sufficient spoken language to meet their daily communication needs (Rose et al., 2016). Consequently, in the absence of targeted and effective intervention, these individuals may reach adulthood with limited communicative autonomy, leading to persistent challenges in social interaction, employment opportunities, and community participation (Light et al., 2019).

Finally, the functional use of language is closely intertwined with the development of social–emotional competencies. Interventions targeting social–emotional functioning aim to address core deficits associated with ASD, with the goal of enhancing both social understanding and emotional regulation (Nuske et al., 2024). Children with ASD frequently exhibit impairments in social–emotional processing, including difficulties in emotion recognition and in understanding and responding appropriately to social interactions (American Psychiatric Association, 2013). These deficits in social–emotional reciprocity can affect both emotional expression and broader patterns of social engagement (Lyu et al., 2024). Promoting inclusion remains a central objective in efforts to enhance social skills, as inclusive environments provide children with opportunities to participate meaningfully, engage in the learning process, and pursue individualized developmental goals (Papazoglou et al., 2021).

ASD and Portable Devices

Children on the autism spectrum appear to demonstrate a distinct preference for digital technologies, a tendency that has been attributed to their affinity for structured, predictable, and discrete forms of information processing (Terlouw et al., 2020). However, this should be understood as a potential preference for the organization and accessibility of digital media rather than as a general preference for digital communication over face-to-face social interaction. The integration of technological tools into educational contexts has expanded the range of intervention approaches available to students with developmental disabilities, with digital technologies potentially serving as feasible platforms for delivering targeted instructional interventions. (Alzyoudi & Alghazali, 2025).

Interventions targeting communication and social–emotional development through the use of computer-based technologies are becoming increasingly prevalent, with a growing body of evidence indicating positive outcomes. These digital interventions have been investigated as potential tools to support core developmental domains, although reported outcomes vary across intervention types and study designs (Ramdoss et al., 2012). Empirical findings further underscore the potential of such approaches to improve both communication and social–emotional skills in children with ASD (Boyd et al., 2015; Garcia-Garcia et al., 2022; Jain et al., 2012).

Portable devices, including mobile phones and tablets, constitute a central component of contemporary digital technologies due to their distinctive affordances that enhance usability and accessibility. These affordances include portability, compact design, multimodal information presentation, and intuitive interfaces. These digital interventions have been investigated as potential tools for supporting core developmental domains, although reported outcomes vary across intervention types and study designs (Vlachou & Drigas, 2017).

Moreover, portable devices facilitate communication through the use of visual supports, such as symbols, images, and written text. Research suggests that children with ASD show a marked preference for such modalities (Aydin & Diken, 2020), likely due to their responsiveness to structured and visually organized environments (Travers & Fefer, 2017). Within this context, tablet-based interventions—particularly those utilizing iPads—have been shown to significantly enhance communication abilities, enabling children to express their needs more effectively and to participate more meaningfully in social interactions (Dundon et al., 2013).

Furthermore, evidence indicates that the use of such devices may contribute to a reduction in communication-related challenging behaviors, primarily through the enhancement of functional communication and social engagement, thereby supporting broader social development (Xin & Leonard, 2015). However, it is essential that the implementation of these technologies be guided by clearly defined criteria and tailored to the individualized needs of each learner, thereby enhancing their effectiveness (Achtypi & Papoudi, 2024).

Beyond serving as platforms for intervention delivery, digital technologies may also influence how users process and respond to information. From the perspectives of technological cognition and transactive technical cognition, mobile-device interventions may involve not only child–device interactions but also socially distributed processes involving caregivers, parents, teachers, or therapists. This distinction is important because outcomes observed in adult-supported interventions cannot necessarily be attributed solely to the digital interface.

Previous reviews have addressed related but distinct aspects of mobile technology use in individuals with ASD. Leung et al. (2021) synthesized randomized controlled trials of mobile-device interventions targeting social and cognitive skills individuals with ASD of different age groups, with the included literature extending to 2019. Hussain et al. (2021) focused on educational mobile applications for children with ASD, particularly their use by teachers to support language and communication skills. More recently, Toma et al. (2024) examined extended-reality-based mobile applications for children with ASD from a predominantly technical and information technology (IT)-oriented perspective, with a specific focus on AR, VR, and MR technologies. Notably, some of the included applications were evaluated only by developers rather than directly with children. Taken together, these reviews suggest the value for an updated synthesis of recent mobile phone- and tablet-based interventions addressing social and communication-related outcomes in children across multiple intervention modalities and heterogeneous research designs.

The present review offers a complementary contribution by providing an updated, child-focused synthesis of studies published between 2020 and 2025, focusing specifically on interventions in which smartphones or tablets served as the primary intervention delivery medium. It integrates contemporary intervention approaches delivered through mobile devices across heterogeneous research designs and outcome domains. Methodologically, the review applies complementary quality appraisal approaches using the MMAT and the SCED Scale, allowing methodological characteristics to be evaluated according to the design-specific requirements of each study type. The review also extends previous work by systematically examining the social validity of the included interventions. In addition, the review reports available information regarding financial support of the included studies to improve transparency regarding potential contextual factors influencing the evidence base. Furthermore, by identifying which interventions employed commercially available applications, the review provides a practical overview of tools that can be accessed and considered by educators, practitioners, and caregivers, subject to their evidence base, suitability, and the individual needs of each child.

2. Methodology

The present review employed a systematic search strategy and was conducted and reported in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. The corresponding PRISMA checklist was completed and is provided in Appendix B. In accordance with PRISMA guidelines, the review was guided by the following focused question: “Do interventions using portable devices enhance social skills in children with ASD?” The review was not registered, and a formal review protocol was not prepared.

2.1. Search Process

A systematic search was conducted across four electronic databases: (a) PsycINFO, (b) Education Resources Information Center (ERIC), (c) Scopus, and (d) Web of Science. Database research was conducted iteratively across PsycINFO, ERIC, Scopus, and Web of Science between summer 2025 and February 2026. The eligibility period was restricted to studies published through December 2025. Exact database-specific search dates were not recorded contemporaneously. These databases were selected to reflect the interdisciplinary yet predominantly educational and psychological focus of the review. ERIC provides specialized coverage of educational research, whereas PsycINFO indexes literature across psychology and the behavioral and social sciences. Scopus and Web of Science were included to provide broader multidisciplinary coverage spanning education, psychology, social sciences, health-related research, and technology. This combination of databases has also been used in previous systematic reviews addressing assistive technology, disability, and autism-related educational outcomes. Databases with a predominant biomedical or technical-computing focus were not searched because the present review aimed to synthesize child-focused social and communication interventions rather than biomedical treatment or technology-development research.

Titles and abstracts of all retrieved records were screened for relevance. The search strategy incorporated the following terms: “Autism” AND (“apps” OR “mobile” OR “tablet” OR “iPad” OR “mobile devices”) AND “social skills”. Studies published between January 2020, and December 2025 were considered eligible, to ensure the inclusion of the most recent evidence in the field of autism research. The selected time frame also complements previous reviews that included earlier technological interventions by focusing specifically on contemporary mobile-device applications. The review process was conducted rigorously and involved critical evaluation by all authors. No automation tools were used at any stage of the review.

The search strategy was intentionally designed to identify studies in which mobile phones or tablets constituted the principal intervention delivery platform. Consequently, intervention-specific approaches such as Augmentative and Alternative Communication, Speech-Generating Devices, Social Stories, Augmented Reality, Serious Games were considered eligible only when implemented through mobile devices and when social skill outcomes satisfied the predefined eligibility criteria.

2.2. Inclusion and Exclusion Criteria

This study aimed to examine the effects of mobile application–based interventions on the social skills of children with ASD. The inclusion criteria for eligible studies were defined as follows: (a) participants had a formal diagnosis of ASD; (b) participants did not present with comorbid conditions (e.g., Tourette syndrome, visual impairments, hearing loss, attention-deficit/hyperactivity disorder [ADHD]); (c) studies were published in peer-reviewed scientific journals or conference proceedings; (d) participants were involved in an intervention incorporating at least one mobile application; (e) the intervention explicitly targeted social skills; (f) studies were published between January 2020 and December 2025; (g) participants were between 2 and 12 years of age.

Studies were excluded if they: (a) were not published in English; (b) did not involve the use of at least one mobile application; or (c) included participants outside the specified age range; (d) pilot studies; (e) were books and reviews. The focus on recent, peer-reviewed publications written in English was intended to capture recent, rapidly evolving evidence. The selected age range was chosen to capture interventions targeting preschool and primary school populations, as intervention approaches for adolescents may differ substantially and warrant separate investigation. Three pilot studies were excluded from the systematic review due to significant methodological and operational limitations. Lee et al. (2023) was omitted due to technical constraints during data acquisition and evaluation. Specifically, storage limitations precluded the execution of the primary protocol for biometric data collection via webcams and wearable devices. The resulting absence of webcam timestamp data restricted the depth of the subsequent analysis. Drapalik et al. (2022) was excluded because its internal validity was compromised by sparse demographic data and a high attrition rate, the latter of which was particularly pronounced within the control group. Safi et al. (2022) presented compromised intervention consistency, as participant absenteeism resulted in multiple missed sessions. Furthermore, adaptations to the research design mandated by the COVID-19 pandemic introduced confounding variables and diminished overall methodological rigor.

For the purposes of this review, the broad outcome domain of social skills constituted an eligibility criterion. Studies were considered eligible when the intervention explicitly targeted social skills and when mobile phones or tablets served as the intervention delivery medium. Studies that did not target social skills or in which mobile phones or tablets were not the intervention delivery medium were not considered eligible for inclusion.

The initial database search yielded a total of 335 records (see Figure 1). Following the removal of duplicates, 224 records remained. Of these, 20 full-text articles were assessed for eligibility. A total of six studies were excluded for the following reasons: three were pilot studies (Drapalik et al., 2022; Lee et al., 2023; Safi et al., 2022); one did not involve the use of a portable device (Alkinj et al., 2022); one included participants with both ASD and ADHD (J. K. Y. Chan et al., 2024); and one did not implement an intervention targeting social skills (Tuba Ceyhun & Kara Savaş, 2025). Consequently, 14 studies were included in the final synthesis.

Figure 1. PRISMA flowchart. **: No automation tools were used; all records were excluded by humans (see the tables in Appendix A).

The search results obtained from the four databases were screened to determine study eligibility according to the established inclusion and exclusion criteria. Title and abstract screening were followed by full-text assessment of potentially eligible studies. The initial screening and selection of studies were conducted by the first author, and the eligibility decisions were subsequently reviewed by the other authors. Any uncertainties or disagreements regarding study eligibility were discussed among the authors and resolved through consensus.

To minimize potential bias and ensure consistency throughout the review process, the established inclusion and exclusion criteria were applied systematically during study selection. Data extraction and coding were conducted by the first author using the predefined data extraction framework and were subsequently reviewed by the other authors. The extracted information and coding decisions were discussed among the authors to ensure consistency and agreement across the included studies.

Agreement was reached at the title and abstract screening, full-text assessment, and data extraction and coding stages, and no unresolved discrepancies remained. Because agreement was established through review and consensus rather than through parallel independent duplicate assessment, no formal inter-rater reliability coefficient was calculated.

2.3. Data Extraction and Coding Measures

To facilitate interpretation and organization, the fourteen studies included in this review were systematically examined, classified, coded, and evaluated manually according to a set of predefined criteria: (a) year of publication; (b) participants’ age range (in years); (c) sample characteristics (including sample size and gender); (d) duration of the intervention; (e) measurement system; (f) the mobile applications and portable devices employed in each intervention; (g) intervention outcomes; (h) reported limitations, along with recommendations for future research; (i) financial support.

When study-level information was not reported or unclear, it was recorded as “not reported”, and no imputation was performed. Descriptive summaries for variables with missing information, such as participant sex, were calculated using only the available reported data. No statistical data conversions were performed.

Because the included studies employed heterogeneous research designs and outcome metrics, no common effect measure was imposed across studies. Where primary studies reported numerical effect-size estimates for outcomes included in the synthesis, these estimates were extracted and retained in the metric reported by the original authors. No effect-size conversions or post hoc effect-size calculations were performed when such estimates were not reported. Moreover, no formal assessment of risk of bias due to missing results, including publication bias or selective reporting, was conducted.

Following data extraction, the specific social skill outcomes reported across the included studies were grouped inductively into three broader outcome categories: (1) communication and pragmatic language skills, (2) social–emotional and interpersonal skills, and (3) cognitive and pre-academic skills. This categorization was developed after examination of the extracted outcomes to facilitate the organization and narrative synthesis of the findings and was not used to determine study eligibility or selection.

2.4. Systematic Visual Analysis of SCED Studies

Intervention effects reported in ten of the fourteen studies included in the review, namely the ten studies employing single-case experimental designs (Alzyoudi & Alghazali, 2025; Chapin et al., 2022; Costescu et al., 2025; Garcia et al., 2023; Genc-Tosun et al., 2023; Güler & Erdem, 2022; Khoirunnisa et al., 2025; Osos et al., 2021; Sng et al., 2022; Waddington et al., 2023) were examined using systematic visual analysis, following the guidelines provided by Kratochwill et al. (2010) in the What Works Clearinghouse Single-Case Design Technical Documentation. The graphical data presented in the original studies were evaluated with respect to level, trend, variability, immediacy of effect, overlap between baseline and intervention phases, and consistency of data patterns across comparable phases, participants, and target behaviors, where applicable. These characteristics were considered collectively to determine whether the graphical data provided evidence of an intervention effect.

No sensitivity analyses were conducted to assess the robustness of the synthesized results.

3. Results

3.1. Participant’s Characteristics

Across the 14 studies included in this review, the total sample comprised 201 children, including 154 children with ASD and 47 typically developing children who served as the active control group in Hassan et al. (2025). Participants’ ages ranged from 3 to 12 years. Sex was not reported in three studies (Anam et al., 2025; Hassan et al., 2025; Khoirunnisa et al., 2025). Therefore, sex-specific descriptive summaries were based only on studies that reported this information. Among the studies reporting sex, data were available for 78 participants, of whom 61 were boys and 17 were girls.

3.2. Duration of the Intervention

The duration of interventions varied considerably across the included studies. In four studies, intervention length was not predetermined but was contingent on participants reaching a predefined mastery or performance criterion (Genc-Tosun et al., 2023; Güler & Erdem, 2022; Osos et al., 2021; Waddington et al., 2023). Five studies reported a defined intervention or use period ranging from one week to two months (Alzyoudi & Alghazali, 2025; Anam et al., 2025; Hassan et al., 2025; Hesami et al., 2025; Khoirunnisa et al., 2025). The remaining five studies described intervention exposure primarily in terms of the number or frequency of sessions rather than a fixed multi-week intervention period (Chapin et al., 2022; Costescu et al., 2025; Garcia et al., 2023; Lyu et al., 2024; Sng et al., 2022). Reported individual session durations varied across studies, although session duration was not consistently reported for all interventions.

3.3. Research Design

The studies included in this review (Table 1) exhibited substantial variability in research design. The majority employed single-case experimental designs. Specifically, three studies used a multiple probe design (Chapin et al., 2022; Osos et al., 2021; Sng et al., 2022), whereas four studies implemented a multiple baseline design (Alzyoudi & Alghazali, 2025; Genc-Tosun et al., 2023; Khoirunnisa et al., 2025; Waddington et al., 2023). Two studies adopted an alternating treatment design (Costescu et al., 2025; Güler & Erdem, 2022), and one study utilized functional analysis (Garcia et al., 2023).

Table 1. This table presents the year of publication, age range, sample, duration of intervention, methodology, measurement system, applications used in the interventions, outcomes, limitations and financial support.

Article & Year	Age	Sample	Duration	Methodology	Measurement System	Application	Outcomes	Limitations	Funding
Osos et al. (2021)	3–5	4 (3 boys-1 girl)	5 to 8 min every day until they reach the goal.	Multiple probe design across participants	Direct observation of behavior using a 3-component task analysis with binary scoring per component; summarized into a 0–3 score per session	Video-Enhanced activity schedules using Apple keynote iPad application	Video-enhanced activity schedules can be an effective way to teach preschoolers with autism to independently start social interactions with their peers	Variability in participant performance, limited assessment of social competence and small sample size	Not reported
Garcia et al. (2023)	7–8	2 (boys)	3 to 5 sessions were conducted daily, 1 to 5 days per week. Each session lasted 5 min	Functional analysis	Caregivers were interviewed and completed QABF by Matson and Vollmer	“Voice meter pro” app through iPhone	For the first child, the intervention was effective, but for the second child, the results were inconclusive and variable; initial step toward measuring and quantifying prosody (e.g., tone, volume) behaviorally	Inability to isolate the boys’ voices from ambient sounds; reduced generalization to other environments; possible punishment effects may have decreased the behavior rather than reinforcing it; the intervention included multiple components making it unclear which element was responsible for the behavior change; limited sample	Not reported
Costescu et al. (2025)	10–12	3 (1 boy-2 girls)	Twice per week; approximately 20 min per session; 5–6 sessions per intervention condition	Single-case experimental design using an alternating treatment design	The three-point rubric evaluated the child’s comprehension answers, while Cohen’s Kappa verified the reliability of treatment fidelity	Social story delivered via digital device through an image display mobile phone application	All participants improved in understanding social norms for friendship, but no significant differences were found between Social Stories delivered traditionally and via mobile devices	Use of identical stories across participants; lack of practical or observational assessment of targeted skills and use of Kohs block design test as the sole measure of cognitive ability	Supported by Babeș-Bolyai University, Grant No. 35285/18.11.2020.
Alzyoudi and Alghazali (2025)	7–8	2 (boys)	Twice per week, 15–20 min each for 7 weeks	Multiple baseline design	Systematic observation and researcher-designed checklist	iPad speech-generating device application	Enhanced functional communication abilities, specifically improving participants’ capacity to initiate requests and express gratitude	Small sample size, conducted exclusively in a school environment and did not provide specific details on how the iPad’s application was used	The authors received no financial support
Anam et al. (2025)	4–7	12 (does not include the gender)	Guided sessions (first 2 days): 3–4 times at least 20 min, independent use (remaining 25 days): not specified times and minutes	Quasi-experimental design	10 close-ended statements rated on a 5-point Likert scale and 9 open-ended questions administered to children’s parents and teachers pre-and post-intervention	“FatimAR” smartphone application	Improvement in letter recognition and recall, social engagement and communication were enhanced	Control group did not include most of the intervention took place at home limiting control over usage conditions and consistency; all data collected through children’s parents and teachers	Open access funding provided by Asia Pacific University of Technology & Innovation (APU)
Hesami et al. (2025)	4–9	30 (23 boys and 7girls)	20 min sessions, every other day for two months	Empirical pre-test–post-test design with a two-month follow-up; participants were randomly divided into an experimental group (ABA + Kookism) and a control group (ABA)	MacArthur Bates Communicative Development Inventories	Serious Game “Kookism” through mobile phone	The ‘Kookism’ serious game produced a statistically significant enhancement in receptive lexicon	Variation in autism severity may limit generalization; measurement bias because of parent-reported data; presence of echolalia in some participants and short follow-up duration	The authors received no financial support
Genc-Tosun et al. (2023)	5	2 (boys)	3 days per week, with a maximum of two 10 min sessions per day, until mastery	Multiple baselines across participants design	The percentage of correct independent multi-step responses across 10 trial sessions were recorded. Data reliability was assessed through interobserver agreement (IOA), while treatment integrity was measured using fidelity checklists. Outcomes were analyzed using visual inspection of graphed data and Tau-U effect size statistics.	“Dokun Konus” iPad-based speech-generating device	Improvements in question-answering skills	Small sample size, discrete trial teaching in home settings, clinical settings only for one participant, and limited research on SGDs for complex communication skills	The authors received no financial support
Waddington et al. (2023)	5–9	2 (1 boy and 1 girl)	Once per week 4 to 20 trials per session until mastery	Multiple baselines across participants design	Percentage of independent requests selected on an iPad-based speech-generating device to continue social routines with parents and to choose between routines, with reliability assessed through inter-observer agreement.	“Proloquo2Go” app through iPad	Children showed improvement in requesting continuation of actions during a variety of social routines with their parents	Small sample size, limited ability to demonstrate a functional relationship, no follow-up phase, and limited assessment of generalization	New Zealand Government Marsden Fund, administered by the Royal Society of New Zealand, Grant No. 10-VUW-071; Victoria University of Wellington; University of Canterbury; New Zealand Institute of Language, Brain & Behavior
Chapin et al. (2022)	3–5	3 (2 boys and 1 girl)	2–3 times per week for 5 to 10 intervention sessions	Multiple-probe design across participants	Interobserver agreement through independent video-based observational coding of communicative turns, using point-by-point agreement	Visualization of video content through “EasyVSD” app via a Samsung Pro tablet, Model SM-T900	All participants demonstrated an increase in the number of communicative turns	Small sample size, limited number of intervention sessions, no follow up tests, and gaps in data collection and program scheduling	Rehabilitation Engineering Research Center on Augmentative and Alternative Communication (The RERC on AAC) from the US Department of Health and Human Services, National Institute on Disability, Independent Living, and Rehabilitation Research [NIDILRR grant #90RE5017 and #90REGE90014]; U.S. Department of Education [#H325D090042]
Khoirunnisa et al. (2025)	7–12	4 (does not include the gender)	Six sessions within 1 week, 20 to 30 min each	Multiple baseline designs	Behavioral observation (frequency and accuracy), video-recorded sessions, ratings compared by two researchers with discrepancies resolved through discussion, and teacher interviews.	AR-based PECS mobile application	Improvement in communicational skills by enhancing students’ answering and commenting skills	Small sample size, short duration, no control group, potential observer bias, and limited outcome scope	PMDSU Scholarship under Kementerian Pendidikan Tinggi, Sains, dan Teknologi
Hassan et al. (2025)	5–11	107 children, 60 diagnosed with ASD and 47 typical (does not include the gender)	8 weeks, consisting of structured weekly training (120 min per week) combined with daily practice sessions (at least 15 min per day)	Randomized Controlled Trial (RCT)	Griffith Empathy Measure (GEM) and multiplayer module evaluations	“Zirkus Empathico 2.0” serious multiplayer game mobile game for android	Improvements in enhancing empathy, collaboration emotional understanding and social engagement	Reliance on parent and teacher reports, lack of blinded observer during the assessment; simultaneous therapies in a few persons may have influenced results	Not reported
Lyu et al. (2024)	6 years old mean age (does not include details for every participant)	24 (21 boys and 3 girls)	1 session approximately 40 min for each child	Between-subject controlled experimental design comparing the “EMooly” intervention with a traditional baseline condition with pre- and post-test	Pre- and post-emotion-recognition quizzes, caregiver questionnaires (Likert scale), and semi-structured interviews.	“EMooly” tablet-based game	Improvements in social–emotional recognition	Small sample size, limited realism due to use of cartoon characters, limited duration, risks of bias or misunderstanding in AI-generated content.	This work is supported in part by the NSERC Discovery Grant (2020–03966), Waterloo. AI’s AI4SocialGood grant, and NSSFC Art Grant (22CG184)
Güler and Erdem (2022)	6–7	3 (1 boy and 2 girls)	20 min each session until mastery	single-subject alternating treatments design comparing animated and non-animated mobile social story maps	The measurement system consisted of recording the percentage of correct responses and assessing reliability through interobserver agreement and application reliability measures.	“Mobile social story Map” application through Tablet	Improvements in social communication and listening comprehension skills	Small sample size, narrow age range, no cognitive or developmental assessments for participants, non-optimal intervention settings, unequal session time, and single-observation opportunity time	The authors received no financial support
Sng et al. (2022)	8–10	3 (boys)	Teaching sessions were conducted daily; the number of sessions varied across participants and intervention phases; session duration was not reported.	Multiple probes across participants design	Number of correct on-topic responses selected during iPad probes and on-topic responses during natural conversations, with reliability assessed through interobserver agreement	“Conversation coach” app by Silver Lining Multimedia through iPad	Different results for every participant, only two of them were improved with additional teaching, the iPad alone was insufficient for teaching on-topic response discrimination, requiring more explicit strategies	Small sample size, baseline may be inaccurate for the first participant, and no blind partners	Not reported

Regarding group-based research designs, three studies employed group experimental designs. One study implemented a randomized controlled trial (RCT) (Hassan et al., 2025), another used a between-subjects controlled design (Lyu et al., 2024), and one adopted a quasi-experimental design (Anam et al., 2025). Finally, Hesami et al. (2025) employed an empirical pre-test–post-test design with a two-month follow-up. The authors reported that participants were randomly divided into experimental and control groups.

3.4. Financial Support

Regarding the funding of the included studies, six studies reported their funding sources (Anam et al., 2025; Chapin et al., 2022; Costescu et al., 2025; Khoirunnisa et al., 2025; Lyu et al., 2024; Waddington et al., 2023), four stated that they received no financial support (Alzyoudi & Alghazali, 2025; Genc-Tosun et al., 2023; Güler & Erdem, 2022; Hesami et al., 2025), and four did not report whether they received funding or identify any funding source (Garcia et al., 2023; Hassan et al., 2025; Osos et al., 2021; Sng et al., 2022).

3.5. Quality Appraisal

Methodological quality was assessed using the Mixed Methods Appraisal Tool (MMAT) version 2018, a validated framework developed by the team led by Pierre Pluye at McGill University. The MMAT was selected due to its capacity to critically appraise a wide range of study designs within a single integrated framework.

Twelve of the included studies (Alzyoudi & Alghazali, 2025; Anam et al., 2025; Chapin et al., 2022; Costescu et al., 2025; Garcia et al., 2023; Genc-Tosun et al., 2023; Güler & Erdem, 2022; Khoirunnisa et al., 2025; Lyu et al., 2024; Osos et al., 2021; Sng et al., 2022; Waddington et al., 2023) employed a quantitative non-randomized design. All twelve studies reported clearly defined research questions, and their data collection procedures were aligned with the stated study objectives.

Regarding sample representativeness, one study (Anam et al., 2025) was rated as unclear, as the sample consisted exclusively of children with mild learning needs, which may limit generalizability to the broader target population. With respect to measurement appropriateness, three studies (Alzyoudi & Alghazali, 2025; Garcia et al., 2023; Sng et al., 2022) were rated as unclear, as data collection was conducted exclusively within school-specific contexts, thereby potentially limiting external validity.

In addition, two studies raised concerns regarding measurement procedures. Khoirunnisa et al. (2025) relied primarily on researcher-based observations, while Anam et al. (2025) used parent and teacher reports as primary data sources. Furthermore, two studies were considered to have methodological limitations in implementation. Chapin et al. (2022) included an insufficient number of intervention sessions, whereas Güler and Erdem (2022) reported that several sessions were delivered solely by the researcher, which may have introduced potential bias.

With regard to outcome data completeness, three studies were rated as uncertain. Both Anam et al. (2025) and Khoirunnisa et al. (2025) lacked a control group and implemented relatively short intervention periods, while Chapin et al. (2022) reported incomplete data due to inconsistent participant attendance. Regarding control of confounding variables, Garcia et al. (2023) employed a multi-component intervention design, which limited the ability to isolate the specific contribution of individual components to observed behavioral changes. Finally, intervention fidelity was adequately maintained across all included studies.

The remaining two studies (Hassan et al., 2025; Hesami et al., 2025) employed quantitative randomized designs. Both studies presented clearly defined research questions, and their data collection procedures were aligned with their stated objectives. Hesami et al. (2025) reported that participants were randomly divided into experimental and control groups. However, the specific randomization procedure was not described. However, in terms of baseline comparability, Hesami et al. (2025) reported variation in autism severity across groups, which may have affected initial group equivalence. Regarding outcome data completeness, Hesami et al. (2025) included a relatively short follow-up period, limiting the generalizability of findings over time. Moreover, outcome assessors were not blinded in either study. Hassan et al. (2025) relied on parent and teacher reports, whereas Hesami et al. (2025) collected data through family feedback. Finally, intervention adherence was high across both studies, with participants following assigned protocols as intended.

For reasons of methodological rigor, all studies employing single-case experimental designs were additionally assessed using the Single-Case Experimental Design (SCED) Scale (Tate et al., 2008), as general critical appraisal tools (e.g., MMAT) are not sufficient for evaluating internal validity and experimental control in single-case experimental designs.

To evaluate the methodological quality of the included studies, the 11-item SCED Scale, developed by Tate et al. (Tate et al., 2008), was utilized. This tool is recognized as a reliable and valid instrument for assessing the rigor of single-case experimental designs and fostering higher standards in research reporting. The scale consists of one item describing participant characteristics and ten items specifically addressing methodological quality. These items evaluate fundamental components, including the operational definition of target behaviors, the appropriateness of the experimental design, the quality of baseline data, and the graphical presentation of raw data. Furthermore, the scale assesses critical criteria such as inter-observer reliability, evaluator independence, and the application of statistical analyses to mitigate bias. The SCED Scale comprises 11 items; however, Item 1 (Clinical history) is reported descriptively and does not contribute to the methodological quality score. Accordingly, the total methodological quality score ranges from 0 to 10 and is calculated based on Items 2–11. Item 10 (Replication) of the SCED Scale refers to replication of treatment application and outcomes across participants, therapists, or settings and should not be interpreted as equivalent to the more stringent requirements for repeated demonstrations of experimental effect used by Kratochwill et al. (Kratochwill et al., 2010) to establish a functional relation. The results are presented in Table 2.

Table 2. Review of Articles Included According to the SCED Scale.

Reference	1. Clinical History	2. Target Behaviors	3. Design	4. Baseline	5. Sampling Behavior During Treatment	6. Raw Data Record	7. Inter-Rater Reliability	8. Independence of Assessors	9. Statistical Analysis	10. Replication	11. Generalization	Total Score
Osos et al. (2021)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	No	8/10
Chapin et al. (2022)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	10/10
Sng et al. (2022)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	Yes	9/10
Alzyoudi and Alghazali (2025)	Yes	Yes	No	Yes	Yes	Yes	Yes	No	No	Yes	No	6/10
Genc-Tosun et al. (2023)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	10/10
Waddington et al. (2023)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	9/10
Khoirunnisa et al. (2025)	Yes	Yes	No	Yes	Yes	Yes	No	No	Yes	Yes	No	6/10
Costescu et al. (2025)	Yes	Yes	No	Yes	Yes	Yes	Yes	No	Yes	Yes	No	7/10
Güler and Erdem (2022)	Yes	Yes	No	Yes	Yes	Yes	Yes	No	No	Yes	Yes	7/10
Garcia et al. (2023)	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	No	8/10

3.6. Systematic Visual Analysis of Intervention Effects and Cross-Study Synthesis

Following the methodological quality appraisal using the SCED Scale (Tate et al., 2008), we conducted a systematic visual analysis of the graphical data reported in the ten studies employing single-case experimental designs (SCEDs), followed by a cross-study synthesis of intervention effects. Specifically, Osos et al. (2021) showed low and stable baseline responding, followed by clear increases in the targeted social skill after intervention introduction across all four participants. Effects were immediate and showed limited overlap for Earl and Alex, whereas Milly and Oscar demonstrated more gradual and variable improvement. Collectively, the findings support a replicated intervention effect, although the relative advantage of the video-enhanced over the electronic schedule varied across participants.

Similarly, Chapin et al. (2022) demonstrated low and stable baseline responding, followed by immediate increases in communicative turns after the staggered introduction of the intervention across all three participants, with no overlap between baseline and intervention data. Although intervention responding was variable, particularly for Bella and Noah, the change in level was consistently replicated across participants while those remaining in baseline continued to show minimal responding. Nevertheless, the graphical data support a functional relation, although the 27-day gap before Noah’s intervention reduces confidence in the final replication.

In contrast, the findings exhibited by Sng et al. (2022) were less consistent. Visual analysis showed low and relatively stable baseline responding, with no clear improvement following the original iPad intervention across the three participants. Kurt improved only after changes to the probe procedure and additional prompting, Albert improved following explicit teaching with laminated cards, and Leo showed no intervention effect. Taken together, the effects were inconsistent across participants and cannot be attributed specifically to the iPad app, providing no convincing evidence of a functional relation for the original intervention.

Alzyoudi and Alghazali (2025) displayed stable baseline performance followed by an immediate increase and a consistent upward trend during intervention for both participants, with little overlap between phases. However, the intervention appears to have been introduced simultaneously for only two participants, rather than being staggered across at least three tiers. Therefore, the graphs suggest improvement following the iPad-based intervention; however, the available graphical evidence does not satisfy accepted standards for demonstrating a functional relation according to the Kratochwill et al. criteria.

Moreover, Genc-Tosun et al. (2023) showed low and stable baseline responding, followed by clear increases after the staggered introduction of the intervention across three question categories for both participants. Although some effects were delayed and initially overlapped with baseline, particularly for Cem, responding eventually reached high and relatively stable levels, with the effect replicated across behaviors. Collectively, the graphical data support a functional relation, although variability and procedural modifications were observed for some target behaviors.

Waddington et al. (2023) also demonstrated low baseline responding followed by immediate increases in correct requests during intervention for both participants, with little overlap between phases. Grace’s responding was initially variable before stabilizing at high levels, whereas Jeremy showed a gradual upward trend after an initial increase, with both participants maintaining high performance during the choice phase. Although the multiple-baseline design included only two participants, the findings indicate positive intervention effects; however, the graphical data provide insufficient evidence to demonstrate a functional relation according to the Kratochwill et al. (2010) criteria.

Khoirunnisa et al. (2025) exhibited increasing baseline trends followed by further gradual improvements during the AR-based intervention, with little overlap between phases across participants and outcomes. However, the intervention was not clearly staggered across participants, and performance remained high or continued to increase during the second baseline phase. Therefore, the findings indicate positive changes in answering and commenting skills, although the available graphical evidence remains insufficient to demonstrate a functional relation according to the Kratochwill et al. criteria.

Güler and Erdem (2022) used an alternating-treatments design to compare animated and non-animated mobile social story maps. For social communication outcomes, baseline responding was low and stable for all three participants, followed by clear increases after intervention introduction. Arda reached criterion after three animated and four non-animated sessions, Hanife after six and eight sessions, respectively, and Zübeyde after two sessions in both conditions. High responding was generally maintained during follow-up, and generalization was demonstrated for Arda and Zübeyde and partially for Hanife. Nevertheless, the two treatment conditions converged at high-performance levels, and the graphical data did not demonstrate a consistent differential effect favoring animated over non-animated delivery. Thus, the study provides clear evidence of improvement relative to baseline but limited evidence that one presentation format was superior to the other.

Costescu et al. (2025) also employed an alternating-treatments design, comparing printed and mobile Social Stories. Baseline performance was stable with low intra-individual variability, and all three participants showed immediate increases in social-norm comprehension after intervention introduction, with little or no overlap between baseline and intervention scores. However, the mobile and printed conditions overlapped substantially and produced similar levels and trends. The authors’ Wilcoxon analyses showed significant differences between baseline and each intervention condition for all participants, but no significant differences between mobile and printed delivery. Therefore, the visual and statistical evidence supports improvement following Social Stories but does not support a differential advantage of mobile delivery.

Garcia et al. (2023) conducted a multielement functional analysis followed by a reversal-design treatment evaluation of function-based differential reinforcement combined with visual feedback from a smartphone app. For Sam, loud voice-volume responses decreased substantially when the modified DR of A-VVR + Q-VVR intervention was introduced, increased during the return to baseline, and decreased again when treatment was reintroduced, supporting a replicated treatment effect. For Damien, loud responses initially decreased during treatment and increased when baseline was reinstated; however, the decrease was not reproduced to the same extent after treatment reintroduction, and responding remained variable even after an enhanced intervention was introduced. Accordingly, the treatment evidence was clear for Sam but mixed/inconclusive for Damien. Moreover, because the smartphone app formed part of a multicomponent function-based intervention, the observed changes cannot be attributed specifically to the app.

Beyond the visual evidence of intervention effectiveness, several studies demonstrated limitations regarding data analysis and generalization. Osos et al. (2021), Sng et al. (2022), Alzyoudi and Alghazali (2025), Güler and Erdem (2022), and Garcia et al. (2023) did not report formal inferential statistics or single-case effect-size estimates, whereas Costescu et al. (2025) supplemented the visual analysis with Wilcoxon tests. Furthermore, limitations regarding generalization were evident across several studies. Osos et al. (2021) and Alzyoudi and Alghazali (2025) did not systematically program, measure, or report the generalization of communication skills across novel settings, untrained individuals, or naturalistic environments. Similarly, Costescu et al. (2025) did not directly assess the application of learned social norms in naturalistic interactions, while Garcia et al. (2023) did not provide evidence of generalization or maintenance across settings and time. In contrast, Güler and Erdem (2022) explicitly assessed generalization, with evidence of transfer reported for two participants and partial generalization for a third.

Methodological constraints concerning observer bias and assessor independence were also evident across several studies. In research conducted by Khoirunnisa et al. (2025), Waddington et al. (2023) and Alzyoudi and Alghazali (2025), the data assessors were internal to the research and intervention teams without explicit blinding to the study conditions. Güler and Erdem (2022) reported high inter-observer and application reliability but did not report assessor blinding or independence for the primary outcome assessment, while Costescu et al. (2025) reported inter-rater reliability without explicit blinding of study personnel. Garcia et al. (2023), in contrast, employed a second independent observer and reported high interobserver agreement across functional-analysis and treatment sessions. Notably, Khoirunnisa et al. (2025) reported that two researchers compared their observational ratings and resolved discrepancies through discussion and stated that the validity and reliability of the measuring instrument had been tested using SPSS. However, no quantitative inter-rater reliability coefficient or agreement percentage was reported. The study also provided no direct evidence of skill transfer to real-life social interactions. Conversely, Chapin et al. (2022) and Genc-Tosun et al. (2023) demonstrated the strongest methodological rigor among the included SCED studies, satisfying all assessed SCED Scale criteria.

Taken together, across the ten SCED studies, the visual evidence for intervention effectiveness was heterogeneous. Chapin et al. (2022) and Genc-Tosun et al. (2023) provided the strongest evidence of a functional relation, with clear and replicated changes across multiple tiers, while Osos et al. (2021) also demonstrated a clear overall intervention effect, although comparative outcomes varied across participants. Güler and Erdem (2022) and Costescu et al. (2025) showed clear improvements relative to baseline, but their alternating-treatments comparisons did not demonstrate a consistent differential advantage of one delivery format over the other. Garcia et al. (2023) demonstrated a replicated treatment effect for one participant, whereas the effect for the second participant was less consistent. In contrast, Alzyoudi and Alghazali (2025), Waddington et al. (2023), and Khoirunnisa et al. (2025) reported positive changes without sufficient evidence of experimental control according to the criteria applied, whereas Sng et al. (2022) showed mixed and participant-specific effects. Therefore, the findings indicate generally positive but heterogeneous evidence across the ten SCED studies, with substantial variation in the strength of experimental control and consistency of effects across designs and participants.

3.7. Effects of the Intervention

Across the 14 studies included in the present review, a broad range of social skill outcomes was assessed among children with ASD. The studies were categorized according to the type of intervention employed, with portable devices serving as a central component of the intervention delivery. Specifically, two utilized social stories, four involved Augmentative and Alternative Communication, three employed Augmented Reality, two used serious games, one targeted on-topic conversational responses, one utilized video-enhanced activity schedules, and one focused on voice volume responses.

Numerical effect-size estimates were extractable for six of the 14 included studies. Among the SCED studies, Chapin et al. (2022) reported Tau-U = 1.00 for all three participants, Genc-Tosun et al. (2023) reported Tau-U values ranging from 0.37 to 1.00 across participants and question categories, and Waddington et al. (2023) reported Tau-U values ranging from 0.98 to 1.00 across participants and intervention phases. Khoirunnisa et al. (2025) reported Cohen’s d values ranging from 1.76 to 2.47 for question-answering and from 1.62 to 2.67 for relevant commenting, while Hesami et al. (2025) reported partial η² = 0.236 for the post-test group effect and 0.077 at follow-up. Anam et al. (2025) reported one-sample Cohen’s d values for questionnaire ratings relative to the neutral-scale midpoint; these values reflected deviations in questionnaire ratings rather than pre–post intervention effect sizes. The remaining eight studies did not provide numerical effect-size estimates for the synthesized outcomes that could be extracted without additional calculation.

3.7.1. Portable Devices and Social Stories

Social Stories represent an evidence-based instructional approach widely used to support children and adolescents with ASD in the acquisition of appropriate social skills. They function as structured narratives that model effective interpersonal behaviors, thereby promoting socially adaptive responses and facilitating social inclusion (Riga et al., 2021). Two included studies examined Social Stories delivered through portable devices.

The study by Güler and Erdem (2022) compared animated and non-animated mobile Social Story Maps for listening comprehension and two targeted social communication behaviors, greeting and asking for help. Participants had basic expressive language skills and were able to follow verbal instructions. All three had been diagnosed with atypical autism, two were non-readers, while the third was beginning to acquire reading and writing skills. All three participants reached the response criterion for the targeted social communication behaviors under both conditions. Arda reached criterion after three animated and four non-animated sessions, Hanife after six and eight sessions, respectively, and Zübeyde after two sessions in each condition. Follow-up and generalization data were also collected for the targeted social communication behaviors. The intervention was delivered through the “Mobile Social Story Map”, a tablet application developed and implemented for individuals with ASD at a special education and rehabilitation center in a city in Northwestern Turkey. Sessions were conducted in the classroom whenever the participant’s teacher was available. In cases where the teacher had conflicting lecture schedules and could not attend, the researcher conducted the sessions independently in the meeting room.

Similarly, Costescu et al. (2025) compared the effects of traditional (print-based) and digital Social Stories in teaching social norms to children with ASD. Participants had IQ scores ranging from 73 to 85 and mean utterance lengths of approximately three to six words. All participants exhibited difficulties with social interaction and friendship formation. In this study, Social Stories were delivered via a mobile phone using an image-based application. All evaluation and intervention sessions were implemented by trained psychologists in a therapeutic setting. The study reported improvements in participants’ understanding of social norms under both delivery formats, with no statistically significant difference between digital and print-based conditions.

3.7.2. Portable Devices and Augmentative and Alternative Communication

The primary aim of Augmentative and Alternative Communication (AAC) systems is to support children with ASD by addressing both verbal and non-verbal communication challenges that may arise (Lorah et al., 2015). Advances in technology have facilitated the development of tools such as portable devices, which enable the use of symbols, images, and text to support interactive communication tailored to the needs of children with ASD (Alzrayer & Banda, 2017). The integration of AAC systems with portable technologies provides enhanced mobility and usability, while such devices are generally considered socially acceptable and widely available in everyday contexts (Still et al., 2014). Four included studies examined AAC interventions delivered through portable devices.

The study by Alzyoudi and Alghazali (2025) reported increases in participants’ initiation of requests and verbal interactions following an AAC intervention delivered through an iPad-based speech-generating application. Neither participant had functional speech; one also had an intellectual disability and speech impairments, with severe limitations in both expressive and receptive language. The authors suggested the integration of such technologies into special education settings to address complex communication needs. Notably, this intervention was implemented using an iPad-based speech-generating device (SGD) commercial application for the first time with participants from non-Western contexts. Data was collected in the participants’ natural classroom environment, in collaboration with their teacher, to maintain their regular routines.

In a study by Chapin et al. (2022), the Visual Scene Display (VSD) approach was used in conjunction with AAC systems. In this intervention, photographs of meaningful events were displayed on a tablet, with specific regions (hotspots) programmed to activate speech output when touched. The study was conducted using the “EasyVSD” commercial application on a Samsung Pro tablet with three toddlers. All three participants had complex communication needs and used minimal speech that was insufficient to meet their everyday communication needs. The study reported increases in communicative turn-taking and communication engagement during the intervention. The intervention was conducted by the first researcher, a certified special education teacher. The study took place at an early childhood education program that served both children with and without disabilities.

Speech-generating devices (SGDs) constitute a subcategory of AAC systems. These electronic tools incorporate a large set of symbols, including photographs and color or black-and-white line drawings representing words or phrases, and can produce auditory output through digitized or synthesized speech (Schlosser & Koul, 2015). Two included studies examined SGDs delivered through portable devices.

Waddington et al. (2023) conducted a study using the “Proloquo2Go” commercial application on an iPad and examined instructional procedures for teaching children with ASD to use SGDs to request participation in social routines. Specifically, the study evaluated an intervention designed to teach two children—who had already acquired requesting skills for tangible items—to request continuation of social interactions using an SGD. Both participants exhibited substantial receptive and expressive communication delays, with VABS-II age-equivalent scores ranging from 0 to 14 months in receptive communication and 9 months for expressive communication. The social routines included, for example, peek-a-boo, making funny faces, foot massages, singing “Five Little Monkeys Jumping on the Bed,” piggyback rides, tickling, singing “Row, Row, Row Your Boat,” and bouncing on a ball. The study reported increases in participants’ requests for continuation of social routines with their parents and in their selection between different routines. All sessions took place in a university-based clinic. The intervention was primarily delivered by children’s mothers, with implementation support provided by the first three authors throughout the study.

The study by Genc-Tosun et al. (2023) investigated the effects of simultaneous SGD use with dynamic screen displays on the acquisition of multistep question-answering skills. The intervention employed the “Dokun Konus” a commercial iPad-based speech-generating application with two children with ASD. Both participants had very limited functional speech and social skills. One did not use words to communicate whereas the other used approximately five words consistently. Their while cognitive levels could not be determined because neither participant completed the Stanford–Binet assessment. The study reported increases in participants’ correct responses to multistep questions, with maintenance data collected at a five-week follow-up. Teachers also reported perceiving the intervention as effective and feasible despite having no previous experience with SGDs. The intervention with the first child was implemented by a special education teacher, while the intervention with the second child was carried out by a special education doctoral student (third author). For the first child, all sessions were held in his bedroom at home. For the second child, all sessions were conducted in a classroom in the developmental support unit of a university. Three included studies examined AR-based interventions delivered through portable devices.

3.7.3. Portable Devices and Augmented Reality

Augmented Reality (AR) refers to a technology that integrates virtual elements into the real-world environment in real time. Specifically, it overlays digitally generated features onto physical contexts through three-dimensional visualization techniques (Doulah et al., 2023; El Shemy et al., 2024; Marto et al., 2019). AR integrates virtual elements into real-world environments and has been applied in interventions involving children with ASD (King et al., 2014).

The study by Khoirunnisa et al. (2025) investigated an AR-based intervention targeting social communication skills in children with ASD. The sample consisted of four children aged 7 to 12 years with limited or absent expressive language abilities. Communication profiles varied across participants, with difficulties including minimal or limited verbal communication, initiating social interactions, and maintaining eye contact. The findings indicated improvements in communication skills, particularly in participants’ ability to respond appropriately and engage in dialogue. More specifically, following the intervention, children demonstrated increased capacity to answer questions and provide contextually relevant comments within an AR-mediated interactive environment. The intervention was delivered via a custom mobile application. The research was conducted in NASOM, Melaka, Malaysia, in a room adapted for the intervention session by the researchers.

Similarly, the study by Anam et al. (2025) aimed to support children with ASD and mild learning difficulties in learning the English alphabet, while also examining whether the intervention influenced social engagement and communication skills. The study reported improvements following the use of the “FatimAR”, a smartphone application developed using the open-source MindAR framework and the React Native programming language, in combination with the “We Live by Knowledge Book”, in both letter recognition and recall of the English alphabet. In addition, parents and teachers reported increased social engagement and communication attempts among the participating children. The intervention sessions were conducted within the participants’ homes.

Finally, the study by Lyu et al. (2024) introduced “EMooly,” a tablet-based game designed to engage caregivers of children with ASD through the integration of Augmented Reality (AR) and Artificial Intelligence (AI). The intervention aimed to enhance both social skills and empathy in children with ASD. The system was developed collaboratively by five experts in neuroscience and education and incorporated individualized social stories targeting emotion recognition and facial expression interpretation. The intervention was conducted across two special education centers. In addition to the research team, the children’s caregivers and, in some cases, their parents were actively involved in the process. Results from the sample of 24 children indicated improvements in emotion recognition skills following the intervention. All participants had sufficient cognitive abilities to engage in basic verbal communication with their caregivers and to understand basic emotions.

3.7.4. Portable Devices and Serious Games

Serious games constitute a category of educational interventions that differ from entertainment-oriented games due to their explicit design focus on skill acquisition and development. Their primary aim extends beyond recreation, targeting learning outcomes and cognitive enhancement in users (Carneiro et al., 2024). In the context of rapidly evolving digital technologies, serious games offer a range of advantages, including the ability to simulate ecologically valid, real-life scenarios (Simões et al., 2018). Accordingly, they represent a valuable adjunct to conventional therapeutic and educational interventions for individuals with ASD. Two included studies examined serious games delivered through portable devices.

The study by Hesami et al. (2025) focused on the development of a custom mobile serious game titled “Kookism” and evaluated its effectiveness in teaching Persian vocabulary to children with ASD. Participants had verbal comprehension ages ranging from 16 to 30 months, children with severe intellectual impairments or psychological disorders were excluded, although autism severity varied across participants and some exhibited echolalia. Participants were randomly divided into two groups, one receiving Applied Behavior Analysis (ABA) alone, and the other receiving ABA in combination with the ‘Kookism’ game. The children played the game at home while being monitored by their families, and data regarding verbal perception was collected with the assistance of the parents. The results indicated a statistically significant difference between the experimental and control groups in favor of the intervention group with regard to gains in receptive vocabulary. Specifically, participants receiving ABA in combination with the “Kookism” game showed greater gains in receptive vocabulary than participants receiving ABA alone.

Similarly, Hassan et al. (2025) developed a serious multiplayer mobile game, “Zirkus,” for Android devices, aimed at enhancing socio-emotional functioning in children with ASD. Participants demonstrated average-range verbal and nonverbal cognitive abilities and represented heterogeneous range of ASD profiles, including childhood autism, Asperger syndrome, atypical autism, and PDD-NOS. The game incorporates interactive learning modules designed to facilitate emotion understanding and recognition. The primary objective of the study was to improve emotional recognition, empathy, and prosocial behavior. The intervention relied on a collaborative framework involving caregivers and researchers in three different autism centers. Findings indicated that participants demonstrated improved abilities in identifying and differentiating emotional states.

3.7.5. Portable Devices and Additional Intervention Models

Three additional studies were included in this review, as they did not meet the previously defined thematic criteria.

The first study by Sng et al. (2022) examined whether the commercial iPad application “Conversation Coach” by Silver Lining Multimedia could support children with ASD in selecting appropriate, on-topic responses during conversational exchanges, and whether these skills could generalize to naturalistic environments. The intervention employed scripted prompts with three response options: one correct (on-topic and socially appropriate), one partially appropriate, and one off-topic. Moreover, the intervention was carried out at the school by one of the researchers who is a trained special educator. The findings indicated that instruction delivered exclusively via iPad was insufficient for participants to master the targeted conversational skill. Participants had IQ scores ranging from 51 to 75 and all fell within the severe symptom range for ASD on the CARS-2; PPVT age-equivalent scores ranged from 3.1 to 3.8 years.

The second study by Osos et al. (2021) investigated the effectiveness of video-enhanced activity schedules delivered via the Apple Keynote iPad commercial application in improving social skills in children with ASD. All participants used at least two-word phrases, with MSEL age-equivalent scores ranging from 24 to 40 months for expressive language and from 17 to 46 months for receptive language. All sessions took place within the EIBI therapy room during regular therapy hours. The room was physically arranged in a manner like a preschool classroom. The intervention was implemented by the first author, while trained research assistants who regularly worked with the children were present during sessions to collect data. The intervention was compared with electronic activity schedules consisting solely of static images and text. Results indicated variability in learning outcomes as two out of four participants acquired the targeted skills more rapidly using video-enhanced schedules, one participant demonstrated equivalent performance across both conditions, and one participant met learning criteria only under the static image condition. Although video preparation required additional time, differences in participants’ performance were observed across the two instructional conditions.

Children with ASD often experience difficulties in speech prosody, particularly in relation to vocal intensity, rhythm, rate, and intonation patterns. Lastly, Garcia et al. (2023) implemented an intervention targeting excessively loud speech in children with ASD. The participants were two boys who exhibited elevated vocal intensity and vocal stereotypies. Both participants could produce one- to three-word utterances and had basic functional verbal communication skills. The intervention utilized the “Voice Meter Pro” commercial application on an iPad, which provided real-time visual feedback for vocal intensity (blue indicating quiet speech, green appropriate intensity, and red loud speech). All sessions were conducted in the participants’ homes. Results showed that for one participant, the intervention reduced high-intensity vocalizations and increased appropriately modulated speech. For the second participant, outcomes were inconclusive.

Across the included studies, outcomes were categorized into three broad domains: (1) communication and pragmatic language skills, (2) social–emotional and interpersonal skills, and (3) cognitive and pre-academic skills. Findings across these domains were mixed. While several studies reported improvements in targeted behaviors, the available evidence was insufficient to determine whether these improvements represented broader changes in social competence or generalized social functioning.

3.8. Social Validity

Only four out of the fourteen reviewed studies (28.5%) reported data regarding social validity, while the remaining studies did not address this aspect. In all studies that reported social validity data, stakeholders evaluated the interventions and their outcomes positively. Specifically, Güler and Erdem (2022) collected social validity data from the participants’ parents and teachers at the special education center to capture the perspectives of those familiar with the research process. In their study, Costescu et al. (2025) ensured social validity by tailoring the social stories used in the intervention to meet each child’s individual needs. Furthermore, Chapin et al. (2022) assessed social validity by having four early childhood special education professionals, two teachers, an occupational therapist, and a paraeducator, view one-minute video clips of the participants during both baseline and intervention sessions. All four professionals strongly agreed that the intervention was enjoyable for the children and contributed to improving their communication skills. Finally, Genc-Tosun et al. (2023) collected social validity data indicating that even teachers without prior experience using SGDs perceived them as effective and feasible tools for teaching communication skills.

3.9. Generalization and Maintenance Across Studies

Across the 14 included studies, generalization was explicitly measured in five studies: Genc-Tosun et al. (2023); Waddington et al. (2023); Chapin et al. (2022); Sng et al. (2022); and Güler and Erdem (2022). In Lyu et al. (2024) and Hassan et al. (2025), transfer to real-life or untrained contexts was incorporated into the intervention design but was not assessed separately. The remaining seven studies—Costescu et al. (2025), Hesami et al. (2025), Alzyoudi and Alghazali (2025), Osos et al. (2021), Khoirunnisa et al. (2025), Garcia et al. (2023) and Anam et al. (2025)—reported neither programmed nor measured generalisation. Finally, maintenance or follow-up data were collected in only four studies: Hesami et al. (2025); Genc-Tosun et al. (2023); Güler and Erdem (2022); and Hassan et al. (2025). No maintenance data were reported in the remaining ten studies.

4. Discussion

In contemporary contexts, technological development is advancing at a rapid pace in response to evolving human needs. In the present systematic review, portable devices served as the primary delivery medium across a range of intervention modalities, including Social Stories, AR, serious games, AAC, and other related intervention approaches. This widespread adoption is partly attributed to the documented preference of children with ASD for digital technologies Across the included studies, improvements were reported for a range of directly targeted outcomes; however, these findings varied across studies and research designs and should not be interpreted as evidence of overall intervention effectiveness. These findings should be considered within the broader literature on technology-supported interventions for individuals with ASD. (Hussain et al., 2021; Leung et al., 2021; Toma et al., 2024).

An equally significant finding of the present review is the prevalent use of commercial applications in more than half of the included studies (Alzyoudi & Alghazali, 2025; Chapin et al., 2022; Garcia et al., 2023; Genc-Tosun et al., 2023; Hesami et al., 2025; Osos et al., 2021; Sng et al., 2022; Waddington et al., 2023), all of which are easily accessible via major software marketplaces like App Store and Google Play Store. This observation is particularly noteworthy, as it demonstrates that the implementation of these interventions does not require specialized programming skills from the end-user. Consequently, this lowers the barrier to entry, significantly facilitating the adoption of such tools by practitioners in the field including educators, special education teachers, and mental health professionals, thereby rendering this type of intervention more practical and accessible for daily clinical and educational practice.

Regarding study outcomes, improvements were reported in socio-emotional skills (Güler & Erdem, 2022; Hassan et al., 2025; Lyu et al., 2024), communication skills (Alzyoudi & Alghazali, 2025; Güler & Erdem, 2022; Khoirunnisa et al., 2025), as well as in broader domains of social functioning, including social engagement (Anam et al., 2025), social routines (Waddington et al., 2023), and social interaction (Alzyoudi & Alghazali, 2025). In addition, gains were observed in specific components of communicative competence in children with ASD, including receptive vocabulary (Hesami et al., 2025), question-answering (Genc-Tosun et al., 2023), turn-taking (Chapin et al., 2022), prosody (Garcia et al., 2023) and on-topic conversational responding (Sng et al., 2022). Given that several included interventions involved parents or caregivers in delivery of interventions or monitoring, the broader family context should also be considered. Identified getting support, social isolation, disruption of daily routines, anxiety, and emotional outbursts as recurring themes in parental burnout among families of children with ASD (Koçak, 2023).

However, given that this remains a relatively emerging field of research, several methodological limitations can be identified. The most prominent limitation concerns small sample sizes (Alzyoudi & Alghazali, 2025; Chapin et al., 2022; Garcia et al., 2023; Genc-Tosun et al., 2023; Güler & Erdem, 2022; Khoirunnisa et al., 2025; Lyu et al., 2024; Osos et al., 2021; Sng et al., 2022; Waddington et al., 2023), as well as restricted age ranges (Güler & Erdem, 2022), both of which limit the generalizability of findings. From a methodological perspective, several studies lacked long-term follow-up assessments (Chapin et al., 2022; Hesami et al., 2025; Waddington et al., 2023), while others relied heavily on subjective reporting—primarily from parents and teachers—for outcome measurement, which may introduce reporting bias (Genc-Tosun et al., 2023; Hassan et al., 2025; Hesami et al., 2025; Waddington et al., 2023). Finally, certain technical and design limitations were also identified, including variability in ASD severity across participants (Hesami et al., 2025) and limited assessment of social skills within real-world contexts (Alzyoudi & Alghazali, 2025; Garcia et al., 2023). An additional limitation of the present review and the resulting evidence base concerns the exclusion of studies involving children with ASD and co-occurring conditions, including Tourette syndrome, visual impairments, hearing loss, and ADHD. Although this criterion was intended to reduce clinical heterogeneity, it narrowed the representativeness of the synthesized population and limits the ecological validity and generalizability of the findings to the broader and more heterogeneous ASD population encountered in real-world educational and clinical settings. Additionally, the review protocol was not prospectively registered in a public registry, such as PROSPERO, before the review commenced. Consequently, the protocol was not publicly available prior to the start of the review, which may limit the transparency and reproducibility of the review process. The absence of contemporaneously recorded database-specific search dates constitutes an additional limitation, as it reduces the reproducibility of the search process. Although the selected databases provided multidisciplinary coverage, the exclusion of additional clinical and technology-focused databases may have resulted in relevant studies being missed.

Furthermore, study selection and data extraction were not conducted independently by multiple reviewers; therefore, no formal inter-rater agreement statistic was calculated. Although eligibility and data extraction decisions were subsequently reviewed by the co-authors and resolved through consensus, the absence of independent duplicate assessment represents a methodological limitation of the review.

Because of the substantial methodological and clinical heterogeneity across studies, conclusions regarding intervention effectiveness should be interpreted cautiously. Rather than indicating definitive evidence of effectiveness, the present findings identify promising intervention approaches that require replication using larger and methodologically rigorous studies. Related single-subject research also highlights the value of combining systematic instructional comparisons with social validity assessment. For example, Koçak et al. (2023) compared direct instruction with simultaneous prompting to teach concrete concepts to children with ASD, while also evaluating social validity.

Based on the literature evaluated in this review, several recommendations for future research are proposed. Firstly, subsequent studies should prioritize the recruitment of larger participant samples to enhance the external validity and generalizability of outcomes across the wider autism spectrum. Furthermore, given the significant heterogeneity in ASD manifestation, symptom severity, and individualized needs documented across the reviewed papers, future research designs must implement rigorous yet adaptive participant selection criteria to better accommodate these unique population characteristics.

Secondly, the current review mapped literature within a strict five-year window (2020–2025) to focus on contemporary applications. A valuable direction for future secondary research would be to eliminate these chronological boundaries. Expanding the inclusion criteria to encompass a broader chronological range will allow future investigators to amass a more extensive sample of studies, thereby offering deeper insights into these specific intervention modalities.

Finally, ASD encompasses a broad spectrum of clinical manifestations and functional levels, which inherently introduces substantial complexity and variability into empirical research. To maintain a relatively homogenous focus, the present review excluded studies involving participants with comorbid conditions like Tourette syndrome, visual impairments, hearing loss, and ADHD. Consequently, future researchers could expand upon this work by systematically incorporating literature that includes participants with such comorbidities, thereby offering a more representative and ecologically valid overview of these interventions.

5. Conclusions

The studies included in this review employed mobile devices as the primary delivery tool across a range of intervention approaches targeting the improvement of social skills in children with ASD. Across the included studies, improvements were reported in several directly targeted outcomes, suggesting that mobile device-based interventions may help support social skills in children with ASD. However, the available evidence remains heterogeneous and is insufficient to establish their overall effectiveness.

In the context of digital interventions targeting social skills in children with ASD, it is important to distinguish between technology as a delivery medium and technology as an active component of cognition. Many applications are conceptualized as passive tools that simply present social information. However, emerging perspectives on technological cognition (Osiurak et al., 2025) suggest that digital tools can play a more active role by shaping how users process, interpret, and apply social information.

In this regard, the apps reviewed in the present study should not be understood solely as platforms for content delivery, but as interactive systems that scaffold social–cognitive processing. This is particularly relevant for children with ASD, who often experience difficulties in spontaneously integrating social cues. Technology, therefore, may function as a structured cognitive partner that supports the integration of social and technical information, rather than acting as a neutral interface.

It is important to note that, in some of the interventions included in the present review, the use of technology was not limited to direct interaction between the child and the technological tool. More specifically, in several studies, caregivers, parents, or teachers actively supported the implementation of the intervention by providing guidance, monitoring, or assistance during the interaction. In these cases, the use of mobile devices was therefore mediated by another person. From the perspective of transactive technical cognition (Osiurak et al., 2025), such interventions may therefore involve a socially distributed cognitive system in which the child, the supporting person, and the technological tool jointly contribute to the intervention process. Accordingly, when active adult support was part of the intervention, the observed outcomes cannot be attributed exclusively to the digital interface. Consequently, the improvements reported following these interventions do not allow firm conclusions to be drawn regarding the extent to which the acquired skills could be transferred independently of caregiver or teacher support.

More broadly, generalisation was explicitly measured in only five of the 14 included studies, while maintenance or follow-up data were reported in four. Consequently, improvements observed within intervention settings should not be interpreted as evidence of broader generalised social competence or long-term durability, and further research should systematically assess transfer across people, settings, tasks, and untrained behaviours, as well as maintenance over time.

Although methodological quality assessment provides information regarding confidence in the reported findings, higher methodological quality scores should not be interpreted as direct evidence of stronger intervention effects.

This review indicates that mobile technologies are increasingly used as tools within interventions for children with ASD, offering flexibility and accessibility. However, the evidence base is characterized by small and heterogeneous samples, as well as limited follow-up assessments. Publication bias and other forms of reporting bias arising from missing results were not formally assessed and therefore cannot be ruled out. This limitation is particularly relevant given the predominance of single-case experimental designs in the included evidence base, as selective publication or reporting of positive intervention outcomes may have influenced the overall pattern of findings. Therefore, their effectiveness should be interpreted with caution, particularly regarding the transfer of skills beyond the intervention context into real-world social interactions.