Abstract
Keywords
Introduction
Education is a powerful tool for promoting social inclusion and long-term financial independence. Inclusive education enables students with diverse needs to learn together in the same environment. Higher education can significantly impact on the lives of individuals with disabilities by improving their access to the labor market, 1 which is a critical step toward enhancing their socioeconomic status and preventing economic marginalization. 2 However, despite growing efforts to make education more inclusive, students with disabilities remain underrepresented in higher education, even as enrolment trends overall increase. According to Hauschildt et al., 3 approximately 15% of European students report having a disability that impacts their ability to study. Unfortunately, consistent support and accommodations for these students are still lacking in higher education institutions. 2 Without the necessary support and adjustments, their academic performance can be severely impacted, often resulting in poor outcomes 4 and, in some cases, even leading to dropping out. 5
The successful implementation of inclusive education has been greatly supported by technological and strategic developments aimed at improving accessibility, enhancing the learning experience, and ensuring that students with diverse needs receive the necessary accommodations and support throughout their studies. By incorporating AI and other technological advancements, higher education can transform into a truly inclusive space where all students, regardless of their abilities, can thrive. Emerging technologies, particularly artificial intelligence (AI), have significant potential to make education more accessible and customized to individual needs. 6 AI can enable personalized learning pathways for every student, thereby improving both the learning process and outcomes. 7
Although AI-driven educational technologies have demonstrated their potential to enhance inclusivity, there are still significant gaps in our understanding of their impact on students with disabilities. 8 Prior research highlights the role of AI in adaptive learning environments, where systems adjust instructional content based on individual performance. 7 Furthermore, AI-powered assistive technologies, including speech-to-text tools and real-time captioning systems, have been shown to enhance accessibility for students with hearing and visual impairments. 9 However, concerns regarding algorithmic bias and the ethical implications of AI decision-making in education persist. 10 By democratizing education, AI presents promising solutions to overcome barriers to accessibility and inclusion, enabling every learner to reach their full potential. 11 As AI technologies continue to evolve, they are expected to contribute significantly to creating more inclusive learning environments that address the unique needs of diverse learners. These technologies can help personalize education, ensuring that all students, regardless of their abilities, have access to the same opportunities and can realize their full potential. However, despite this optimistic outlook, there is a noticeable lack of research exploring the specific needs of students in actual special educational contexts. 12 Given the inevitable integration of AI into education, 13 it makes sense to better understand its implications and explores how it can be used effectively, without compromising students’ privacy, autonomy, or emotional well-being, ensuring that AI enhances, rather than undermines, the core values of inclusive and ethical education.
The concept of AI in education is still evolving, and there is no universally agreed-upon definition. Much like broader discussions surrounding natural and artificial intelligence, debates persist about the nature of AI and its broader implications. However, there are certain characteristics of AI systems that are particularly relevant when considering their application in education. Typically, AI technologies involve software solutions capable of analyzing their environment and executing algorithms with a degree of autonomy. Key features include the ability to learn independently and display anthropomorphic traits of intelligence, such as “rationality” or “awareness,” at least in specific contexts.
This review explores the potential of AI-driven assistive technologies to promote accessibility, inclusivity, and academic success for students with disabilities in higher education. In contrast to previous reviews that have examined general AI applications, this study delves into the specific impact of AI across various disability types, including sensory, cognitive, and physical disabilities. Additionally, it addresses the ethical considerations associated with the adoption of AI technologies.
Inclusive education and artificial intelligence
Higher education institutions are aiming to make education more accessible, and AI offers significant potential to support this goal. The Universal Design for Learning (UDL) framework recognizes the unique ways in which individuals process information, stay motivated, and express knowledge. 14 As a result, higher education is moving toward a more flexible and personalized learning environment to accommodate the diverse needs of students. Despite the potential of AI to assist students with disabilities, there is a lack of comprehensive analysis on this topic in existing literature. 9 Few studies address the impact of AI on students with disabilities in depth. For instance, while some reviews briefly mention adaptive learning platforms for students with disabilities, they fail to provide a thorough analysis.9,15 Furthermore, AI-based proctoring systems, which monitor students during exams, may increase anxiety, particularly for students with mental health conditions. However, this concern is often overlooked in research. 16
While previous reviews6,10 have explored the role of AI in education, they predominantly focus on general student populations rather than addressing the specific needs of students with disabilities. These reviews primarily explore AI’s potential to increase engagement, automate assessments, and improve personalized learning for all students, but lack a detailed analysis of how AI technologies specifically impact accessibility and inclusivity for learners with disabilities.
Our findings are consistent with existing literature on AI-driven personalized learning 17 and reinforce evidence that AI-powered adaptive systems can tailor educational experiences to individual needs. However, this study extends previous research by highlighting key gaps in AI accessibility, ethical concerns, and institutional barriers that have been largely overlooked. Specifically, while AI tools are widely recognized for their ability to personalize learning experiences, few studies have critically examined their effectiveness for students with disabilities, particularly those with cognitive or neurodevelopmental differences.
Several previous reviews9,18 discuss AI-powered tools such as adaptive learning systems and chatbots. However, these studies do not critically examine whether these tools effectively accommodate students with different disabilities. Furthermore, there is limited research on the potential biases in AI models and how these biases may inadvertently disadvantage students with disabilities. Our review highlights these gaps and highlights the need for more targeted research on AI accessibility and ethical AI design.
Unlike previous studies that broadly assess the role of AI in higher education, our review specifically examines its impact on students with disabilities, making it one of the few studies to systematically evaluate AI-driven accessibility tools in this context. Our findings provide practical insights for institutions seeking to implement AI in an inclusive way.
AI is increasingly being used in education to perform students learning analytics, which involves measuring, collecting, analyzing, and presenting data about learners and their learning environments to better understand and improve educational outcomes. AI technology can apply all four types of learning analytics (1) descriptive analytics to describe what has already occurred, (2) diagnostic analytics to analyze the reasons behind certain outcomes, (3) predictive analytics to forecast what is likely to happen in the future, and (4) prescriptive analytics to provide recommendations on how to achieve desired results. 19 Unlike traditional methods, AI-driven analytics can process the full range of available data and learn from it autonomously. This allows AI to gain deeper insights and offer personalized learning pathways. Another key advantage is that AI operates in real-time, enabling the immediate detection of challenges that students may encounter during the learning process. 19 This dynamic and adaptive approach makes AI a powerful tool for enhancing educational environments and improving outcomes.
Recent literature reviews highlight that knowledge of inclusiveness in AI-powered learning tools remains limited.6,9,10,18 Ethical concerns regarding the use of AI in education, particularly discrimination, are becoming increasingly urgent. These concerns encompass issues such as inclusivity, bias, privacy, error, and social acceptability. For example, bias in machine learning algorithms can lead to discriminatory practices that hinder access to education, while natural language processing (NLP) models may reinforce negative stereotypes about certain disabilities. 9 A primary goal of educational technology is the shift toward personalized learning. 7 To achieve this, large-scale computer adaptive learning (CAL) systems have been developed. These systems automatically adjust the type or difficulty of instructional materials and practice based on an individual learner’s performance. They also provide diagnostic feedback on areas that need improvement, offer guidance on which skills to prioritize, and recommend resources for skill enhancement. 7 Educational chatbots, or conversational agents, have significant potential for delivering personalized and interactive learning experiences to students. 13 In the context of AI, these technologies can help students with disabilities in higher education by enhancing accessibility, inclusivity, and academic performance. Additionally, these tools can streamline administrative tasks, allowing instructors to focus on more meaningful aspects of instruction and mentorship. 20
In our review we focused on investigating how AI can enhance existing assistive technologies (e.g., speech-to-text, AI-powered screen readers, predictive text, etc.) to improve accessibility for students with physical, visual, auditory, or learning disabilities in higher education. The aim of this research was to determine the impact of AI-powered assistive technologies on academic performance, engagement, and accessibility for students with disabilities in higher education. To explore the role of AI in supporting students with disabilities in higher education, we developed the following research question.
Methods and procedures
This study employs an integrative review methodology, following the framework outlined by Whittemore and Knafl. 21 The integrative review was selected over a systematic or scoping review due to its ability to synthesize diverse research methods (qualitative, quantitative, and mixed methods) into a cohesive thematic analysis. The heterogeneity of research approaches and disciplines involved in the study of AI in higher education, which is still in its infancy, necessitates an appropriate review method that can accommodate such diversity. A scoping review differs from a systematic review in that the former allows for a broader scope by incorporating findings from qualitative case studies, experimental research, and theoretical frameworks, as opposed to the typical focus of a systematic review on a well-defined research question with strict inclusion criteria for quantitative studies. While scoping reviews are often used to map the literature and identify research gaps, they do not provide the critical synthesis required to derive practical and theoretical implications. Consequently, the present study employs an integrative approach, a strategy that has been demonstrated to facilitate a more profound comprehension of the way AI-powered assistive technologies influence accessibility, academic performance, and engagement for students with disabilities in higher education.
This approach allows for the inclusion of diverse research methods (qualitative, quantitative, and mixed methods) and integrates theoretical, empirical, and practical insights. This review followed the six-phase approach outlined by Whittemore and Knafl. 21 First, the theme of integrating AI to support students with disabilities in higher education was identified, as discussed in the introduction, and three key research questions were formulated. Second, inclusion and exclusion criteria were established, focusing on the time frame, research status, and language to ensure the relevance of selected studies. Third, key information to be extracted from the studies, including AI-powered assistive technologies, AI tools, AI’s impact on students’ learning, and best practices, was determined. Fourth, the selected studies were thoroughly evaluated. Fifth, the results were interpreted, offering insights into the role of AI in supporting students with disabilities in higher education. Finally, the body of knowledge was synthesized and presented in this review.
In order to integrate the findings derived from a variety of methodologies, a thematic synthesis approach was employed. A thematic synthesis approach was used to analyze and categorize the extracted data. The themes in this study were not predefined but rather emerged inductively from the data analysis. To identify these emergent themes, we followed an iterative coding process, wherein each selected study was systematically reviewed to extract key concepts related to AI-powered assistive technologies and their impact on students with disabilities. Initial open coding was performed to highlight recurring ideas and concepts. The studies were coded based on their methodological approach (qualitative, quantitative, or mixed methods) and categorized into key themes, namely. The present study employed inductive thematic synthesis, allowing key themes to emerge from the data. These codes were then grouped into broader thematic categories based on patterns observed across the studies. Thematic clusters were refined through multiple rounds of review and discussion among the researchers to ensure accuracy and coherence in theme development. The studies were analyzed in three stages: (I) initial coding: two reviewers independently extracted recurring concepts related to AI implementation; (II) category formation: codes were grouped into broader themes based on AI’s role in education; and (III) final thematic clusters: the most frequent and relevant patterns were synthesized into four major themes: (1) AI-driven accessibility tools, (2) personalized learning experiences, (3) institutional barriers to adoption, and (4) best practices for inclusive AI integration. This structured synthesis ensures that insights from different research paradigms contribute to a coherent understanding of AI’s role in inclusive higher education. The synthesis process followed three key criteria. - - -
(“AI-powered assistive technology” OR “artificial intelligence”) AND (“disability” OR “students with disabilities” OR “special needs”) AND (“higher education” OR “university” OR “college”) AND (“accessibility” OR “academic accessibility” OR “inclusion”) AND (“academic performance” OR “learning outcomes” OR “student achievement”) AND (“inclusivity” OR “inclusive education” OR “inclusive learning”) AND (“learning tools” OR “adaptive learning tools” OR “assistive technology tools”) AND (“speech-to-text” OR “screen readers” OR “text-to-speech” OR “assistive devices”) AND (“adaptive learning systems” OR “personalized learning tools”).
This search string included a broad range of terms related to the key areas of interest in our research and helped capture studies addressing AI-powered assistive technologies in higher education. The search was conducted in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) guidelines.
22
As illustrated in Figure 1, our search strategy included: (1) Initial identification of 1200 studies across four databases (PubMed, IEEE Xplore, Web of Science, Scopus); (2) Screening based on title and abstract, leading to 250 articles for full-text review; and (3) Final inclusion of 27 studies, selected based on predefined eligibility criteria (AI application, assistive technology focus, higher education context). Of the 89 reports sought for retrieval, 56 were excluded at the full-text screening stage. Among these, 51 articles were excluded due to not meeting the inclusion criteria (e.g., irrelevant outcomes, inappropriate study design, lack of focus on AI tools and inclusion), and five were excluded due to lack of access to full text despite repeated attempts through institutional and interlibrary resources. PRISMA flow chart.
We included both quantitative and qualitative studies (peer-reviewed journal articles, conference proceedings, dissertations, and reports) published in English between 2020 and 2024. A search strategy was employed to identify relevant studies, including both quantitative and qualitative research methodologies. This approach included reviewing peer-reviewed journal articles, conference proceedings, and institutional reports. The term “reports” in this context refers to institutional white papers, government publications, and conference findings that offer critical insights into the application of AI in the domain of assistive technologies in education. While some of these reports have undergone a peer-review process, others constitute grey literature, which has been critically assessed for credibility. The adoption of AI in accessibility and education represents a nascent field of study, with various observations documented in institutional reports as opposed to the conventional format of academic journals. These reports offer real-time data on the implementation challenges and emergent solutions pertaining to AI, thereby serving as invaluable resources for analyzing practical applications that have not yet been adequately addressed in academic literature.
Although the primary inclusion criterion focused on empirical studies reporting measurable impacts on student learning, accessibility, or engagement, the review also included select conceptual papers and systematic reviews. These were incorporated when they contributed critical theoretical perspectives, synthesized empirical findings, or offered valuable insights that informed understanding of AI tools and digital literacy in educational contexts.
While the review prioritized empirical studies with clearly reported methodologies and measurable outcomes, a small number of non-empirical sources were included. These were selected based on their conceptual relevance, clarity of argumentation, and potential to enrich the thematic synthesis—particularly in areas where empirical data was sparse or emerging. For example, brief reflective contributions 23 were included when they offered practitioner insights or contextual interpretations that aligned with identified themes. These sources were treated as supplementary and were coded using the same thematic framework as empirical studies, with caution taken to distinguish between opinion-based content and substantiated claims.
In addition to empirical studies, we included a small number of systematic literature reviews (
The present review focuses on studies published between 2020 and 2024, due to the rapid advancements in AI technologies during this period. While the use of AI in education has been a subject of study for many years, studies prior to 2020 often employed AI models that are now considered outdated (e.g., rule-based systems as opposed to deep learning or neural networks). However, key studies providing foundational insights from before 2020 were manually screened and referenced where relevant.
The existence of selection bias is a possibility, given the presence of language limitations and the exclusion of certain grey literature. However, this focus ensures the inclusion of high-quality, peer-reviewed research. Moreover, it is important to note that AI technologies are subject to rapid evolution; therefore, the 2020–2024 time frame encompasses the most recent advancements in deep learning, natural language processing (NLP), and AI-driven accessibility tools.
Inclusion criteria and exclusion criteria.
Following the application of the inclusion and exclusion criteria, a total of 27 studies were selected for this review. The selection process was guided by several key criteria to ensure the relevance, methodological rigor, and empirical value of the included studies. Empirical studies in this review are those that use qualitative, quantitative, or mixed-methods approaches to systematically collect and analyze data related to AI-powered assistive technologies in higher education. This includes experimental research, case studies, surveys, and ethnographic studies. Firstly, relevance was a primary consideration, meaning that only studies explicitly focused on AI-powered assistive technologies in higher education were included. Research that did not specifically address AI-driven tools, or that examined general educational technology without an assistive focus, was excluded.
Secondly, methodological rigor played a crucial role in the selection process. Priority was given to studies that demonstrated a transparent research design, provided clear details about their sample populations, and used replicable methods. This criterion ensured that the findings were reliable and could contribute meaningfully to the broader discourse on AI in education. Additionally, direct AI application was an essential requirement. Articles that examined non-AI assistive technologies, such as traditional screen readers or basic accessibility tools without AI-driven functionalities, were excluded. The present review concentrated on technologies that used artificial intelligence to enhance accessibility, learning experiences, or student engagement.
Studies presenting empirical evidence were considered, with the inclusion criterion being the reporting of measurable impacts on student learning, accessibility, or engagement. The incorporation of empirical data ensured that the findings were based on observed outcomes rather than theoretical discussions or speculative analyses.
Data extraction and analysis
List and description of selected studies.
Based on the inclusion and exclusion criteria, 27 papers were selected for this integrative review (see Figure 1). Key data were extracted from the selected studies, including recurring concepts, applied AI technologies, contexts, results, and conclusions. The extracted information was analyzed to identify and organize recurring themes related to AI integration in higher education. For each identified theme or pattern, a code was assigned. Coherent thematic clusters were created by grouping similar codes based on their relationships and aspects of AI integration.
A thematic synthesis was conducted to categorize and analyze the data into major themes: (1) personalized learning, (2) benefits of AI-driven assistive technology, (3) adoption challenges and institutional barriers, and (4) best practices for AI implementation. This synthesis process involved analyzing and summarizing studies on the integration of AI into inclusive higher education, guided by three key criteria:
Thematic frequency
This criterion focuses on identifying prevalence of themes or concepts across studies through key term analysis of titles, abstracts, and keywords.
AI technology used
This involved examining the specific AI tools, techniques, and algorithms employed to support students with disabilities in higher education.
Results
The findings from each study were reviewed, with particular attention to the integration of AI in supporting students with disabilities and its implications for educational practice.
Given the diversity of study designs (empirical, conceptual, and systematic reviews), a quality appraisal was conducted. Critical appraisal tools, such as CASP for qualitative studies 24 and PRISMA for systematic reviews, 22 were used to assess the quality of the included studies. We evaluated each empirical study for clarity of research aims, methodological transparency, appropriateness of data collection tools, and rigor in analysis. Conceptual and review papers were assessed for relevance, depth of analysis, and theoretical contribution. While formal scoring was not applied, notes on methodological strengths and limitations were taken to inform thematic synthesis. As this integrative review did not involve primary data collection from human participants, ethical approval was not required. However, ethical considerations were carefully observed in the interpretation and reporting of the research data.
Results
This integrative review aimed to explore current AI-powered assistive technologies available and their applications in supporting students with disabilities in higher education. As research on AI and its integration into education continues to grow at an exponential rate, this review specifically focused on studies that addressed AI-powered assistive technologies for students with disabilities in higher education, ultimately selecting 27 relevant studies. The primary goal was to examine how AI is being used to enhance accessibility, personalize learning experiences, and improve academic outcomes for students with various disabilities. Although there is a wealth of studies on AI in education, a significant gap remains in understanding how these technologies are specifically tailored to meet the needs of students with disabilities. This includes areas such as adaptive learning, real-time feedback systems, and AI-driven accommodations for cognitive and physical challenges. Notably, Toyokawa et al. 19 found it challenging to identify studies that explore AI applications for students with disabilities in Japan, highlighting a broader lack of research in this crucial area. The thematic analysis of the selected papers identified four key categories in the applications of AI to support students with disabilities in higher education. The first theme, “AI and personalized learning,” explored how AI can customize educational experiences to meet the unique needs of individual students, promoting more effective and engaging learning for students with disabilities. The second theme, “Benefits of AI-driven assistive technology,” was highlighted in six studies that specifically examined various technologies used to support students with disabilities in higher education. The third theme, “AI challenges and barriers,” addressed the obstacles faced in the adoption of AI-powered assistive technologies in higher education. Finally, “Best practices” theme focused on examples of AI-powered tools that have successfully addressed the needs of students with different types of disabilities.
Comparative summary of key AI applications in higher education for students with disabilities.
When integrating AI into education, the student or learner should be placed at the core of the process, surrounded by various educational technologies and systems designed to enhance learning outcomes. Ahmad et al.
26
highlighted several AI applications in education that address student diversity and foster inclusive universities. These include: (1)
The reviewed papers suggest several AI-based assistive technologies that have significantly enhanced the learning experience for students with disabilities in higher education: (1)
The review of the selected articles revealed several key findings regarding the impact of AI on higher education, particularly in areas such as personalization, efficiency, accessibility, and early intervention, while also addressing associated challenges. Among the empirical studies included the study by Yunusov et al. 30 offers one of the most detailed quantitative analyses of AI-driven adaptive learning systems, reporting an 85% improvement in student learning outcomes Furthermore, educators reported a 70% increase in student engagement when AI tools were integrated into the curriculum.
Other studies suggested that AI-driven assessments helped reduce grading time by an average of 50%, while maintaining accuracy levels comparable to traditional methods. 30 However, educators raised concerns about AI’s ability to effectively assess complex, creative assignments.31–33 AI was also found to strengthen stronger teacher-student relationships by enhancing daily interactions, allowing teachers to focus more on meaningful engagement with students while administrative tasks were streamlined by AI. 20 Additionally, AI positively impacted the digital literacy and computer skills of teachers, promoting increased access to digital teaching resources. 20 AI also played a key role in helping teachers create a more inclusive and accessible learning environment.34,35
A broad range of methodological rigors was observed across the 27 included studies. Most empirical studies (e.g., Refs. 18 and 20) clearly articulated their research questions and presented data collection and analysis procedures transparently. However, several others (e.g., Ref. 30) lacked detail on sampling or validation strategies, limiting replicability. Non-empirical studies showed variability in depth and analytical richness. For example, while Song and Xie 23 provided reflective insights relevant to ITS design, their brevity and lack of methodological structure limit their generalizability. Similarly, the systematic reviews (e.g., Ref. 20) offered synthesized findings but varied in reporting the included studies’ characteristics, complicating our ability to assess thematic frequency precisely. These limitations were considered during synthesis, with greater interpretive weight given to well-documented empirical studies. Nonetheless, all studies contributed conceptual or contextual value to understanding ITS in inclusive education.
Discussion
This review systematically explored the role of AI in higher education, focusing specifically on its impact in supporting students with disabilities. The evidence presented across 27 studies highlights the significant potential of AI to enhance inclusivity, accessibility, and learning outcomes for students with disabilities. By integrating AI into higher education, universities can create more personalized, engaging, and effective learning environments, ultimately fostering inclusiveness. These findings not only align with existing theories of accessibility and inclusive education but also expand upon them. This section discusses the key areas in which AI influences accessibility and inclusivity, as well as its broader implications for both education theory and practice.
Our findings are aligned with those of preceding studies, which have emphasized the role of AI in supporting personalized learning (see Ref. 13). However, the present study also reveals critical challenges that have not been widely discussed in previous literature, such as the lack of AI tools designed specifically for students with cognitive disabilities. While existing research has focused on AI’s potential for improving accessibility, the present study highlights the necessity of incorporating universal design principles into AI-based assistive technologies to ensure equitable learning opportunities for all students.
AI-powered technologies have been increasingly contributed to the creation of personalized learning experiences for students with disabilities. A substantial body of research has been published on this topic, including studies by Panjwani-Charani and Zhai, 36 Song and Xie, 23 and Babo et al. 1 These studies have highlighted the importance of AI-driven adaptive learning systems in tailoring educational content to the individual needs of students, thereby enhancing both engagement and accessibility. For instance, Yunusov et al. 30 found that students using AI-driven adaptive learning tools demonstrated an 85% improvement in learning outcomes compared to traditional methods. Moreover, AI-based tutoring systems, such as MetaTutor, have been observed to enhance students’ metacognitive skills by providing real-time feedback on their learning strategies. 18
Personalized learning
A major theme emerging across the included studies is the capacity of AI-ATs to adapt to learners’ cognitive, sensory, and behavioral needs through personalization features such as adaptive learning pathways (e.g., Ref. 13), predictive content delivery (e.g., Ref. 37), and voice or gesture-based interfaces (e.g., Refs. 6 and 38), which answer our RQ 1. Personalization in education has been a topic of ongoing discussion for several decades, not only in relation to students with disabilities but also as a core element of a student-centered approach.17,39,40 AI’s ability to tailor educational content and adapt to individual needs makes it a powerful tool in promoting inclusivity and enhancing overall learning outcomes. 36 Consequently, AI fosters a more positive and engaging learning environment where students feel valued and recognized. 20 By assessing student behavior, learning patterns, and preferences, AI enables the creation of customized learning pathways that cater to specific disabilities.
AI uses computational methods, algorithms, data analytics, and automation to extract meaningful insights from the vast amount of data generated in educational settings. These insights, referred to as learning analytics, hold the potential to create personalized learning experiences and personalized learning experiences and environments tailored to students’ unique needs and preferences. 26 AI-powered personalized learning systems employ algorithms that make data-driven decisions to optimize the educational experience, offering customized recommendations based on collected learning data.
AI technologies facilitate continuous monitoring of student progress, enabling real-time adjustments to learning environments and course content based on individual preferences and prior experiences.13,41 One of the key challenges in developing personalized learning systems is the creation of diverse educational content that considers the various ways students perceive and process information. For teachers, it is nearly impossible to account for all the factors influencing students’ performance. However, with the support of AI tools in co-teaching, higher education can move closer to offering personalized learning experiences. These AI tools not only tailor educational content to each students’ abilities but also adjust the presentation methods, complexity, and volume to better align with individual learning needs.
Tapalova and Zhiyenbayeva 34 highlight the popular language-learning application, Duolingo, as an example of how AI can create personalized learning experience. This app uses machine learning and natural language processing to tailor the learning process. It starts with an adaptive entry test, tracks user errors, and offers various forms of content interaction, all supported by instant feedback. AI-powered chatbots simulate real-life conversations, immersing learners in situational scenarios that enhance the learning experience. This approach can be especially beneficial for students with disabilities, as they can access the platform as often as needed, ensuring flexible and personalized practice.
In addition to the studies by Johnson et al., 42 several other papers reported adaptive instructional systems as a key feature of ITS. These studies emphasized dynamic content tailoring in response to learner profiles. For instance, Song and Xie 23 discussed reflective design strategies that indirectly supported personalization through teacher involvement.
Benefits of AI-driven assistive technology in supporting students with disabilities
All the selected studies emphasize that AI-driven assistive technologies can significantly improve learning outcomes for both students with and without disabilities (RQ 2). These technologies enhance accessibility and offer diverse learning experiences, presenting a valuable alternative to traditional assistive devices. 18 This approach allows students with disabilities, but not exclusively, to participate in a more flexible and personalized learning environment. 34 In a study conducted at Spanish universities, students with disabilities recognized the potential of educational technologies, praising the wide range of opportunities these tools provide for enhancing their learning experiences. 43
The selection of AI-powered assistive technologies was informed by the prevalence of literature on the subject, empirical evidence of their effectiveness, and their adoption in higher education settings. The review incorporated only tools for which there was peer-reviewed evidence supporting their impact on students with disabilities. The effectiveness of AI-powered assistive technologies in improving accessibility for students with disabilities is highlighted by several studies. For example, Neha et al. 25 found that speech-to-text software significantly improved academic performance among students with hearing impairments. Similarly, Vistorte et al. 18 demonstrated how screen readers enhance accessibility for visually impaired learners, improving content comprehension and engagement.
AI tools are changing how students are trained by providing real-time feedback and valuable insights. 13 However, many educators lack the necessary training to effectively integrate these tools, resulting in a knowledge gap. 20 To ensure successful AI integration in supporting students with disabilities in higher education, prioritizing the professional development of teachers is essential. 20 Research shows that AI can play a crucial role in various aspects of training, such as managing knowledge, assessing needs, organizing training sessions, and giving feedback on outcomes. For example, Intelligent Tutoring Systems like MetaTutor offer feedback on how students’ emotions influence their learning and provide guidance on managing those emotions for improved results. MetaTutor goes further by assessing not only students’ cognitive abilities but also their metacognitive skills, making it a comprehensive tool for enhancing both learning and emotional regulation. 18
AI-driven assistive technology benefits in supporting students with disabilities.
AI-powered assistive technologies have been shown to significantly enhance accessibility and personalized learning for students with disabilities. 8 These tools, including speech-to-text software and adaptive learning platforms, provide tailored support, thereby reducing learning barriers and fostering academic success.46,47
Reported benefits include increased autonomy, reduced reliance on human assistance, and enhanced access to learning materials (e.g., Ref. 23). Challenges, however, were equally prevalent, particularly in terms of data privacy, user trust, and the need for institutional readiness. Systematic reviews in the sample (e.g., Refs. 14, 16, and 17) also noted that technical barriers, such as limited language or gesture recognition accuracy, disproportionately affect users with complex or multiple disabilities. These insights underscore the importance of embedding ethical and usability considerations into AI-AT development.
Challenges in adopting AI-driven assistive technology to support students with disabilities
The theme “Challenges in AI Adoption” has been consolidated into a single section in order to prevent redundancy. This section highlights key issues such as data privacy concerns, teacher preparedness, and AI bias, all of which impact the successful integration of AI-powered assistive technologies in higher education.
Our review identified several challenges in integrating AI technology into higher education to support students with disabilities. As noted in recent studies, Zhang J and Zhang Z 20 emphasize that AI have yet to significantly reduce educational disparities. Other studies point to major obstacles in the successful implementation of AI in supporting students with disabilities. One key challenge is gaining user acceptance and finding an effective balance between AI-assisted learning and traditional teaching methods. 48 Additionally, ensuring robust data privacy and security measures is critical to protect sensitive student information. 20 Another concern is the potential over-reliance on technology, which may undermine the importance of human interaction and mentorship, particularly in teacher training.20,49 Furthermore, excessive dependence on AI could disadvantage individuals with disabilities in the job market, as highlighted by Samuel & Kolawole. 50
Whilst the implementation of AI applications has the potential to enhance accessibility, it must be noted that there are significant challenges associated with this process. These challenges primarily concern the necessity of ensuring equitable access and the minimization of bias. It is therefore vital to understand these barriers if effective adaptive learning environments are to be designed.
AI-generated data may also lead to unequal and discriminatory outcomes. Numerous studies have shown that AI tools, such as ChatGPT, can produce biased outputs that reflect political, religious, racial, gender, or fairness-related biases. 10 According to Alshahrani et al., 10 AI-generated information must be balanced and accurate for all users to prevent the amplification of existing societal inequalities. Despite the accessibility advantages AI brings to education, there is growing concern that it could inadvertently exacerbate existing biases, further disadvantaging certain groups. 51 To address this risk, AI tools must be developed and implemented within an ethical framework that promotes inclusivity and fairness. Ensuring that AI has a positive and equitable impact on all student groups is essential for its successful integration into higher education.10,5,52
AI bias remains a critical concern in education, as machine learning algorithms have the potential to unintentionally reinforce societal inequalities. Research has demonstrated that AI-powered educational tools may disproportionately disadvantage students from underrepresented backgrounds if the training data lacks diversity. 10 Moreover, natural language processing (NLP) models employed in AI tutoring systems have been observed to perpetuate negative stereotypes about disabilities, which raise concerns about their fairness and inclusivity.
AI bias remains a significant ethical challenge in education. For example, speech-to-text AI tools often misinterpret the speech patterns of people with disabilities, resulting in inaccurate transcriptions. 6 Similarly, AI proctoring tools have been found to disproportionately flag students with disabilities for “suspicious behavior,” particularly those with involuntary movements or assistive technologies. 16 These biases risk exacerbating, rather than reducing educational inequalities.
The integration of AI in education raises significant privacy concerns. AI-powered proctoring systems, for example, use facial recognition and keystroke monitoring to prevent cheating, but these methods have been criticized for violating student privacy and disproportionately misidentifying certain demographic groups. 51 Furthermore, AI-driven learning analytics collect vast amounts of student data, often without explicit consent, leading to concerns over data security and potential misuse. In order to minimize risks to student data, AI developers should adopt transparent data collection policies and prioritize privacy-preserving AI techniques, such as differential privacy and federate learning.
While AI can make educational processes more efficient, there is growing concern that excessive reliance on technology could diminish personal interactions between teachers and students. 53 This shift may also hinder the development of essential skills such as critical thinking, creativity, and problem-solving. 10 Additionally, the risk of students using AI tools to cheat presents a significant challenge, potentially allowing them to receive academic recognition without genuinely engaging in learning. This raises important questions about academic integrity and how educational institutions should address these evolving challenges. To maintain high educational standards, it is crucial that AI be used responsibly enhancing learning rather than replacing meaningful engagement.
Another concern revolves around the mental health of both students and teachers. “AI anxiety” refers to the fear and uncertainty surrounding the rapid development of AI, particularly regarding job displacement and its unpredictable effects. 42 Educators are concerned that AI might replace their roles, prompting questions about the future of employment in education and other sectors. 10 Additionally, the high cost and complexity of scaling AI, coupled with the lack of clear guidelines and specialized expertise, pose significant barriers to its effective implementation in education. 10 Addressing these challenges is crucial for ensuring that AI has a sustainable and positive impact on education.
Best practices
To answer to our RQ3, several studies highlighted the value of participatory design and user-centered development (e.g., Refs. 20 and 37). Best practices include multimodal interfaces, customizable settings, explainable AI features, and interoperability with existing learning management systems. Conceptual contributions emphasized the alignment with Universal Design for Learning (UDL) frameworks and iterative prototyping with feedback from diverse disability groups. While the evidence on implementation quality varies, the studies collectively suggest that inclusive design principles must be embedded from the early stages of AI tool development.
While the number of studies on AI applications in education is growing rapidly, there remains a relatively small body of research specifically focused on AI’s role in promoting inclusive education. This gap may be attributed to factors such as “AI anxiety” and teachers’ lack of preparedness to integrate AI in their daily teaching practices. Despite these challenges, our review uncovered several studies that provide valuable insights into the use of AI-driven assistive technologies to support students with disabilities in higher education. For example, Zingoni et al. 41 explore how BESPECIAL, a supporting software platform, helps dyslexic students navigate the challenges they encounter in university settings. Preliminary results indicate the platform’s effectiveness as proof of concept for personalized learning approaches. 41 Another example is the chatbot Sammy, which was tested among 215 undergraduate students to assess its potential as an inclusive tutor. 45 The findings suggest that chatbots can create supportive environments for disadvantaged students by answering questions, connecting them with resources, and accommodating various learning styles through personalized interactions. 45 These examples identify key features that enhance AI usability, such as accessibility, interactivity, and immediate feedback, underscoring the importance of design considerations that prioritize inclusivity.
The integration of AI tools within higher education to support students with disabilities must be approached with a structured and best-practice approach to ensure the equitable, effective and sustainable use of these technologies. The following key principles can be generalized across diverse educational contexts.
Teacher training and AI literacy
The fundamental requirement for effective AI adoption is to equip educators with the necessary skills and knowledge to use AI tools effectively. A significant proportion of educators are not familiar with AI-powered assistive technologies, resulting in underutilization or improper implementation. Training programs should be developed to assist educators in integrating AI into their teaching strategies, thereby making learning more inclusive and adaptive for students with disabilities.
Ethical AI design
AI models must be designed with fairness, transparency, and inclusivity at their core. The presence of biases in AI algorithms can disproportionately affect students with disabilities, limiting accessibility or reinforcing inequalities. It is therefore essential that developers ensure that AI-based assistive tools undergo rigorous bias testing and adhere to ethical AI guidelines to provide equitable opportunities for all students. Clear and explainable AI decision-making processes should be implemented to build trust among users.
User-centered development
It is crucial that AI tools are not designed in isolation; rather, they should be co-developed with direct input from students with disabilities. Engaging users in the design process is essential to ensure that AI-powered solutions align with real-world accessibility needs, preferences, and learning styles. Inclusive design frameworks, such as participatory design and usability testing with diverse student groups, should guide AI development in educational settings.
By prioritizing these fundamental principles, educational institutions have the potential to transcend the mere adoption of AI-powered assistive technologies and instead foster a more inclusive, student-centered learning environment. In this paradigm, AI functions not as a hindrance but as an empowering tool, enabling a more equitable and conducive learning environment for all students. 54
Recent studies have indicated the efficacy of AI-powered assistive technologies in enhancing accessibility for students with disabilities. 54 For instance, Neha et al. 25 discovered that speech-to-text software substantially improved academic performance among students with hearing impairments. In a similar vein, Vistorte et al. 18 demonstrated how screen readers can augment accessibility for visually impaired learners, thereby enhancing content comprehension and engagement.
Ethical AI solutions
To ensure that the benefits of AI-powered tools are maximized and potential risks mitigated, several ethical AI solutions are necessary. A major concern in education is the lack of transparency in AI decision-making processes. Many AI-driven tools, such as automated grading systems and adaptive learning platforms, function as “black boxes,” which makes it difficult for students and educators to understand how decisions are made. Explainable AI (XAI) addresses this issue by enhancing transparency and trust. By rendering the decision-making processes of AI more interpretable, XAI enables users to comprehend the rationale behind specific recommendations or actions. This is of particular significance for students with disabilities, as it ensures that assistive technologies provide clear, justifiable adaptations tailored to their individual needs. Furthermore, XAI can contribute to the alleviation of concerns regarding fairness in grading and learning assessments by offering insights into the generation of automated feedback. Future advancements in XAI should prioritize user-friendly explanations that accommodate diverse cognitive and learning abilities, ensuring accessibility for all students.
Students with disabilities are particularly at risks of AI bias, due to the lack of diverse user input in the training of most models. A notable example of this bias is the tendency of AI-generated captions to misinterpret dysarthric speech, creating accessibility barriers in online learning environments. Furthermore, AI-based essay grading systems may penalize students with cognitive disabilities unfairly if they structure their arguments differently to their neurotypical peers. 9 These findings emphasize the urgent need for inclusive AI training datasets and bias audits in educational AI tools.
Limitations
This review has several limitations in exploring available AI-powered assistive technologies and their applications in supporting students with disabilities in higher education. First, it is clear that very few peer-reviewed studies directly address the intersection of AI, accessibility, and student learning for individuals with disabilities in higher education. Second, by excluding studies focused on primary or secondary education, we may have overlooked valuable research that could offer transferable insights. However, the primary aim of this review was to focus specifically on the unique challenges and opportunities AI presents in supporting students with disabilities in higher education, particularly in fostering more inclusive learning environments. Additionally, this review highlights a lack of understanding of the impact of AI on academic research within higher education institutions. While some studies provide robust methodologies for assessing AI’s influence on student academic outcomes, there is a clear lack of research specifically targeting students with disabilities. To address this, future studies should be funded and conducted to assess the impact of AI on academic performance within the educational context, with a particular focus on higher education settings. This would help to deepen our understanding of how AI can enhance both accessibility and academic success for students with disabilities.
Conclusions
This integrative review explored the potential of integrating AI into higher education to support students with disabilities. The findings show that AI has the potential to significantly enhance student assessment and foster inclusive education. Inclusive technologies can improve learning outcomes for all students. AI-driven technologies offer a wide range of opportunities to innovate and improve higher education practices, particularly in terms of personalization and individualization. AI-based services contribute to the personalization and individualization of the learning process by utilizing educational analytics and data engineering methods. However, it is important to recognize and address challenges and limitations that may arise when implementing these tools.
Footnotes
ORCID iDs
