The rapid advancements in modern technologies have opened new possibilities for enhancing educational experiences for students with Special Educational Needs and Disabilities (SEND). This paper conducts a systematic review of 139 studies on the integration of AI, VR, and LLMs in Special Education. Using a deductive thematic analysis framework, it identifies key themes and challenges to synthesize the current state of knowledge and propose future research directions. The findings underscore the transformative potential of AI and Immersive Technologies in fostering personalized learning, improving social engagement, and advancing cognitive development among SEND students. Additionally, current SEN methodologies and practices are defined, teachers'attitudes toward inclusion and technology adoption, and the prevailing technological tools utilized, based on various syndromes and disorders. Challenges such as ethical considerations, accessibility barriers, and resource limitations are also discussed. Moreover, this study explores tailored technological tools specifically designed to meet the unique needs of these students. The paper concludes with limitations, recommendations for cross-sector collaboration and inclusive policies to ensure these technologies are effectively utilized to prioritize and enhance learning experiences for SEND students, as well as future directions.

Large language models (LLMs) have created remarkable possibilities for analyzing and generating language data and have been integrated into several fields aiming to transform them, including education. While most research efforts focus on LLMs in typical education or social robots, limited applications of LLMs have been reported in special education. Moreover, there is a profound lack of combined research in LLM-based social robots in special education. In this work, we argue that although LLMs and social robots have demonstrated their potential to advance special education separately, their combination is not yet fully exploited, and further research is required to enable such use. The first objective of this work is to review relevant literature to assess the feasibility of developing LLMs on social robot platforms for use in special education. The second objective of this work is to reveal related challenges, limitations, opportunities, and ethical considerations to provide insights, aiming to subsequently formulate guidelines for the efficient integration of LLM-based social robots into special education practices. To this end, the third objective of this work is to propose a thoughtful framework, aiming to formulate a safe and inclusive learning environment for students in special education, suggesting actionable steps that could be followed by educators, developers and stakeholders, towards address the unique needs and challenges of students with diverse learning requirements.

This systematic review examines 205 studies on the use of Large Language Models (LLMs) in psychiatry, psychology, psychotherapy, and clinical workflows. Furthermore, studies that directly evaluated at least one LLM in a mental health context were included in the extended detailed analysis. GPT-4 and GPT-3.5 were the most commonly assessed models. Although LLMs showed promising short-term performance across domains, most evaluations relied on small, non-longitudinal datasets and single-session testing, limiting generalizability. The evidence indicates rapid growth but significant methodological inconsistency, emphasizing the need for more diverse datasets, standardized evaluation, and long-term validation before clinical integration. This review also examines how LLMs are being incorporated into mental health practice, outlining key challenges, limitations, and emerging opportunities. Ethical, clinical, and technological considerations are proposed to guide responsible adoption. Given the complexity of mental health care, a multidisciplinary, human-centered approach remains essential to ensure that future LLM applications augment—rather than replace—professional expertise.