Generative AI in disability-inclusive learning: a bibliometric and systematic literature analysis

Generative Artificial Intelligence (GenAI) has rapidly permeated education, with growing implications for disability-inclusive practice. Objective: This review maps GenAI uses for students with disabilities since public LLM adoption, identifies research clusters, and surfaces gaps. Methods: Following SPAR-4-SLR, we searched Scopus and Google Scholar (publications Jan 1, 2022–Feb 6, 2025; English; journals/conferences). After screening, 88 records were retained for the qualitative SLR; a relevance subset (n = 49, score = 3) underwent bibliometric and text-mining analysis using TF-IDF, K-Means (k = 5), and PCA based visualisation; keyword co-occurrence networks were built in VOSviewer. Results: Five clusters emerged: (1) adaptive tools for autism and language learning; (2) inclusive/early-childhood AI integration; (3) game-based/adaptive learning; (4) broad ChatGPT applications across K–12/higher/special education; and (5) frameworks/conceptual models. ASD and dyslexia dominate; visual/hearing/motor impairments are underrepresented. Conclusions: GenAI shows promise for personalisation, AAC, and teacher support, but evidence is early-stage and uneven across disabilities. Recommendations include standardised reporting (datasets, prompts, guardrails), longitudinal evaluation, and policy frameworks aligning Universal Design for Learning (UDL) with AI ethics and privacy. A replication package (data/code) and a taxonomy for GenAI-inclusive learning are proposed.

Expand GenAI interventions beyond ASD and learning disabilities to address major gaps in rehabilitation support for visual, hearing, motor, and other complex disabilities as well: The current evidence base is disproportionately narrow, signalling an urgent need for inclusive GenAI design that meets the rehabilitation needs of all disability groups.
Position GenAI as a catalyst for personalised, multimodal rehabilitation by integrating text, speech, image, and VR/AR outputs to strengthen Augmentative and Alternative Communication (AAC), social-communication training, motor practice, and real-world simulation: Such multimodal tools can enhance functional engagement and reduce barriers in both clinical and educational rehabilitation settings.
Embed accessibility-first and ethics-centred frameworks such as Universal Design Language (UDL),Web Content Accessibility Guidelines (WCAG) including privacy-preserving architectures and bias mitigation into GenAI development to ensure safety, equitable access, and meaningful participation of neurodivergent and disabled learners: These guardrails are essential for responsible deployment in rehabilitation contexts.
Prioritise longitudinal, co-designed, and multidisciplinary research to validate real-world rehabilitation outcomes, ensure cultural and linguistic relevance, and support scalable adoption especially in low-resource or underserved areas: Sustained evidence generation will enable GenAI tools to evolve from promising prototypes into reliable components of rehabilitation practice.

Expand GenAI interventions beyond ASD and learning disabilities to address major gaps in rehabilitation support for visual, hearing, motor, and other complex disabilities as well: The current evidence base is disproportionately narrow, signalling an urgent need for inclusive GenAI design that meets the rehabilitation needs of all disability groups.
Position GenAI as a catalyst for personalised, multimodal rehabilitation by integrating text, speech, image, and VR/AR outputs to strengthen Augmentative and Alternative Communication (AAC), social-communication training, motor practice, and real-world simulation: Such multimodal tools can enhance functional engagement and reduce barriers in both clinical and educational rehabilitation settings.
Embed accessibility-first and ethics-centred frameworks such as Universal Design Language (UDL),Web Content Accessibility Guidelines (WCAG) including privacy-preserving architectures and bias mitigation into GenAI development to ensure safety, equitable access, and meaningful participation of neurodivergent and disabled learners: These guardrails are essential for responsible deployment in rehabilitation contexts.
Prioritise longitudinal, co-designed, and multidisciplinary research to validate real-world rehabilitation outcomes, ensure cultural and linguistic relevance, and support scalable adoption especially in low-resource or underserved areas: Sustained evidence generation will enable GenAI tools to evolve from promising prototypes into reliable components of rehabilitation practice.