The Impact of Artificial Intelligence on Inclusive Education: A Systematic Review

Abstract

Artificial intelligence (AI) is regarded as a pivotal instrument in the realm of inclusive education, offering a means to enhance accessibility and personalize learning experiences for students with disabilities. This study presents a comprehensive and systematic review of the impact of AI on inclusive education, elucidating both its advantages and the challenges associated with its implementation. In accordance with the PRISMA guidelines, studies published between 2021 and 2024 in databases including Scopus, Web of Science, ScienceDirect, and ERIC were subjected to analysis. A bibliometric analysis was conducted using Bibliometrix to identify key trends, and ATLAS.ti was employed to organize topics such as accessibility, personalization, and ethics. The findings demonstrate that AI enhances accessibility through the provision of adapted materials, including image descriptions for visually impaired students and audio transcripts for those with hearing impairments. Furthermore, it alleviates the administrative burden on educators, enabling them to prioritize pedagogical guidance. Nevertheless, several obstacles persist, including a dearth of AI training, inadequate infrastructure, and ethical concerns regarding privacy and equitable access to technology. Ultimately, AI holds immense promise for enhancing inclusive education and fostering greater accessibility. However, its success hinges on surmounting these challenges. This study underscores the necessity for policies and strategies that ensure the ethical and sustainable utilization of AI in inclusive environments.

1. Introduction

Inclusive education is recognized as a fundamental human right, as established by international frameworks such as the Convention on the Rights of Persons with Disabilities (Naciones Unidas, 2006). It is understood as an ongoing process aimed at transforming educational systems to eliminate barriers that limit student participation and learning, thereby ensuring equal opportunities for all. As noted by Santos and Leal (2023) and Upadhyay (2023), inclusive education promotes environments that value diversity and foster active engagement from every student, particularly those with special educational needs or disabilities (Santos & Leal, 2023; Upadhyay, 2023; Volker et al., 2022). However, the effective and equitable implementation of inclusive education faces challenges in different educational settings, such as a lack of resources, inadequate teacher training, and attitudinal barriers (Allan, 2022; Sijuola & Davidova, 2022; Prathama et al., 2022).

Within this context, artificial intelligence (AI) emerges as a promising tool to help dismantle barriers to inclusive education by enhancing accessibility, personalization, and instructional quality for students with special needs. Here, equity is understood as the fair allocation of resources and opportunities, tailored to the specific learning requirements of each student. This concept involves not only ensuring access but also adjusting pedagogical strategies to accommodate diverse abilities (Volker et al., 2022).

AI has shown a strong potential in adapting educational environments to meet individual learning needs, thereby supporting academic performance and a personalized curriculum delivery (Gibellini et al., 2023; Knox et al., 2019; Woolf et al., 2013; Toyokawa et al., 2023). Intelligent tutoring systems, for example, adjust to students’ pace and learning style, improving comprehension and retention through machine learning algorithms that identify areas of difficulty and deliver targeted feedback (Carbonell Bernal & Hernández Prados, 2024).

Moreover, AI can ease teachers’ administrative workload, allowing more time for pedagogical engagement and high-impact practices (Zahurin et al., 2024). As noted by Khine (2024), AI also facilitates the development of accessible materials, such as image descriptions for students with visual impairments or automatic speech transcriptions for those with hearing impairments—advancing real classroom inclusion.

During the COVID-19 pandemic, the accelerated adoption of digital technologies highlighted the urgent need to acquire new technological competencies in the educational field. In particular, AI has enabled technologies such as intelligent tutoring systems and adaptive MOOCs to foster accessible and personalized learning for students with varying abilities (Toyokawa et al., 2023; Page et al., 2021; Sein-Echaluce et al., 2016). This context also revealed significant technological inequalities, such as the lack of devices and connectivity, which disproportionately affected students with special needs (Prathama et al., 2022).

Recent studies suggest that, if properly implemented, AI can significantly enhance students’ active participation and equity in learning (Toyokawa et al., 2023; Sghaier et al., 2022). However, the introduction of AI in inclusive education also raises ethical and technological challenges, such as data privacy, equitable access to technology, and resistance from some communities to its adoption (Choez Calderón & Miranda Bajaña, 2024; Hong et al., 2018; Klimova et al., 2023). An example of these challenges can be observed in the study by Choez Calderón and Miranda Bajaña (2024), where a school implemented an AI-based intelligent tutoring system that collected sensitive data about students’ performance and needs. Although the tool improved learning personalization, the lack of clear policies regarding data management raised concerns among parents about privacy and the potential misuse of information. This case underscores the need to establish robust ethical standards to protect the rights of students and their families.

Despite its potential benefits, there is a lack of research that comprehensively synthesizes findings on the use of AI in inclusive education. Although Toyokawa et al. (2023) address AI-supported active reading and Tuna (2022) studies humanoid robots for children with autism, there are still no studies that analyze how these technologies could benefit students with multiple disabilities in underserved contexts. This systematic review aims to fill this gap by providing a detailed analysis of the impact of AI on learning accessibility and personalization, while also identifying the main challenges and limitations faced by educational institutions in implementing AI in inclusive settings. Given the emerging nature of this field, this study adopts an exploratory approach to synthesize existing findings and provides a broad understanding of the potential applications and challenges of AI in inclusive education. By taking this approach, this study seeks to address key dimensions of this underexplored topic, identifying trends and gaps to guide future research efforts.

Furthermore, the current practices, benefits, and ethical dilemmas associated with AI use will be evaluated to provide recommendations that can guide future research and inclusive educational policies. This study is structured around three main research questions: How does AI enhance the accessibility and personalization of learning for students with special needs or disabilities? How does AI contribute to reducing teachers’ administrative workload and improving their interaction with students in inclusive environments? And what are the main challenges and limitations faced by educational institutions when implementing AI in inclusive settings?

Ultimately, this review seeks to contribute to the development of a more inclusive educational framework that harnesses the potential of AI to ensure equitable access to quality education for all learners, regardless of their needs. Beyond showcasing the benefits of AI in promoting inclusion, this study also addresses practical and ethical challenges that must be considered for its sustainable and responsible implementation across diverse educational contexts.

2. Materials and Methods

This systematic review was conducted following the PRISMA guidelines (Preferred Reporting Items for Systematic Reviews and Meta-Analyses), in alignment with the recommendations of Moher et al. (2009) and Page et al. (2021). Although PROSPERO registration is typically required for systematic reviews, this study was not registered because its thematic scope—focused on educational practices and artificial intelligence—does not fall within the health and social care domain primarily covered by PROSPERO. These protocols were used to ensure methodological rigor and transparency throughout the stages of the identification, selection, evaluation, synthesis, and reporting of the relevant studies.

The primary objective of this review is to examine the impact of artificial intelligence (AI) on inclusive education, with a focus on students with special educational needs or disabilities.

2.1. Research Questions

Three specific questions were formulated to guide the analysis of the selected studies:

• RQ1. How does artificial intelligence enhance the accessibility and personalization of learning for students with special needs or disabilities?
• RQ2. How does AI contribute to reducing teachers’ administrative workload and improving their interaction with students in inclusive environments?
• RQ3. What are the main challenges and limitations faced by educational institutions when implementing AI in inclusive settings?

These questions aim to capture both the opportunities and the challenges involved in implementing AI in inclusive education, allowing for a comprehensive and balanced assessment of its impact.

2.2. Search Strategy

The search strategy was informed by a preliminary bibliometric analysis using Bibliometrix, an open-source R-package selected for its robust capabilities in scientific mapping and descriptive analysis. This tool facilitated the identification of trends in publication volume (Figure 1), the most frequent author keywords (Figure 2), conceptual clusters (Figure 3), and patterns in country-level scientific output (Figure 4, Figure 5 and Figure 6). The analysis included both descriptive indicators—such as publication trends, keyword frequency, and citation counts by country—and relational techniques. In particular, a co-occurrence network of author keywords was generated to identify thematic clusters (e.g., the strong association between “inclusive education” and “artificial intelligence”), while country-level collaboration patterns were visualized through co-authorship mapping. These outputs contributed not only to understanding the structure of the field but also to validating the scope and relevance of the review itself.

Crucially, this bibliometric phase informed the refinement of the inclusion and exclusion criteria. For instance, the high recurrence of terms such as “inclusive education”, “students with disabilities”, and “artificial intelligence” confirmed that these should remain central anchors in the search string design. Similarly, the keyword clustering and country-specific publication patterns helped to delineate the thematic and geographical focus of this review, reinforcing the decision to exclude articles that lacked a clear empirical link between AI and inclusive educational practices.

Based on these insights, a comprehensive search strategy was constructed using Boolean operators (AND, OR) to combine the key terms. The search was conducted between April and May 2024 in four academic databases: Scopus, Web of Science, ScienceDirect, and ERIC. These databases were selected for their established academic rigor and for offering a broad coverage of peer-reviewed research in education, technology, and interdisciplinary studies. In contrast, repositories such as PubMed and IEEE Xplore—though valuable for biomedical and engineering research—were found to have a limited representation of educational inclusion topics. Google Scholar was excluded due to concerns regarding transparency, replicability, and quality control.

The chosen timeframe (2021–2024) was also guided by bibliometric evidence. Figure 1 shows a clear rise in the scientific output on AI and inclusive education during this period, with a marked peak in 2023. The slight decrease observed in 2024 is attributed to incomplete indexing at the time of data collection. This range coincides with the post-pandemic expansion of AI applications in educational settings and captures the most relevant empirical contributions aligned with the review’s objectives. Earlier studies, particularly those published before 2021, tended to be conceptual or disconnected from the context of students with disabilities.

Figure 2 presents a word cloud that illustrates the most frequently used expressions by authors in the literature on AI and inclusive education, specifically in their keywords. The terms “inclusive education” (22 mentions) and “artificial intelligence” (14 mentions) appear most prominently, followed by “education”. This highlights the centrality of inclusive education and artificial intelligence in current studies, underscoring the relevance of these topics in educational research.

In contrast, Figure 3 illustrates a cluster that closely associates “artificial intelligence” with “inclusive education”, emphasizing the pivotal role of AI as a tool to advance inclusive educational methodologies. This finding not only corroborates the coherence of the key terms utilized in the search but also demonstrates the robust interconnection between both concepts in contemporary research.

This analysis underscores the pivotal role of artificial intelligence in advancing inclusive educational practices, which is vital for the formulation of pedagogical strategies that effectively integrate students with special needs into the teaching–learning process. The identification of these clusters suggests that research in this area is focused on the ways in which artificial intelligence can be leveraged to enhance accessibility, personalization, and equity within educational environments.

Figure 4 illustrates that India and the United States are the foremost contributors to the corpus of scientific articles pertaining to AI and inclusive education, with 27 and 26 publications, respectively. Additionally, Spain and the United Kingdom have demonstrated a notable dedication to this field with 10 and 9 articles, respectively.

Figure 5 emphasizes the importance of international collaboration in this field. The United States, India, and the United Kingdom lead in both national and international research partnerships, while countries such as Canada, Ecuador, and France also show active participation in cross-border initiatives. This collaborative landscape highlights the global relevance of AI for inclusive education and the need for shared strategies and insights.

Figure 6 presents an analysis of citations, which reveals that France and the United Kingdom have the highest number of citations, reflecting the influence of their research. In the fifth position is Ecuador, which has demonstrated a growing interest in the subject matter.

The findings obtained from the bibliometric analysis were used to develop search terms that included combinations of the following: “inclusive education” OR “students with disabilities” AND “artificial intelligence” OR “AI” AND “learning”. Operators were adapted for each database in order to maximize the retrieval of relevant studies. To ensure broad and relevant coverage, specific filters were applied:

• Temporal Range: studies published between 2021 and 2024 were included, given the recent increase in research on AI and inclusive education, particularly following the COVID-19 pandemic.
• Study Types: only primary studies (original research articles, randomized controlled trials, and observational and experimental studies) were included, excluding reviews, unpublished theses, and personal communications, to ensure the empirical validity of the findings.
• Access: the selection was limited to open-access studies, ensuring their availability to the research community.
• Language: Studies in English and Spanish were included, as bibliometric analysis revealed that the majority of scientific production in the field comes from English-speaking countries, with India being notably relevant in the analysis. However, no articles were found in languages such as Hindi or other regional languages of India. Spanish was selected due to its predominance in countries like Ecuador, where the primary language is Spanish, and its notable representation in the scientific literature related to AI in inclusive education.

To enhance transparency and replicability, the Boolean expressions and search syntax were tailored to the requirements of each database.

Table 1. Database-specific search strategies used in the review.

Database	Search Strategy
Scopus	TITLE-ABS-KEY ((“inclusive education” OR “students with disabilities”) AND (“artificial intelligence” OR “ai”) AND (“learning”)) AND PUBYEAR > 2020 AND PUBYEAR < 2025 AND DOCTYPE = article AND LANGUAGE = English or Spanish AND Open Access = all AND Publication Stage = final
Web of Science	TOPIC: (“influence” AND (“artificial intelligence” OR “AI”) AND (“teaching-learning practice” OR “teaching-learning” OR “didactic” OR “didactic design”)) AND Publication Years = 2021–2024 AND Document Type = Article AND Language = English or Spanish AND Open Access = All
ScienceDirect	(“inclusive education” OR “students with disabilities”) AND (“artificial intelligence” OR “ai”) AND (“learning”), with filters manually applied: Year = 2021–2024, Article type = research, Language = English/Spanish, Open Access = Yes
ERIC	(“inclusive education” OR “students with disabilities”) AND (“artificial intelligence” OR “ai”) AND (“learning”), using search interface filters for years, language, and document type

summarizes the structure of the search strategies used.

2.3. Criteria for Inclusion and Exclusion

To ensure the rigor and replicability of the study, specific inclusion and exclusion criteria were established to guide the selection of relevant articles. These criteria were designed to align with the objectives of the systematic review and to maintain transparency throughout the research process.

• Inclusion Criteria
  • Temporal Range: studies published between 2021 and 2024 were included to capture recent advancements in the application of AI in inclusive education, particularly in response to the increasing use of educational technologies following the COVID-19 pandemic.
  • Type of Studies: peer-reviewed original research articles, randomized controlled trials, observational studies, and experimental studies that focused on the application of AI in inclusive education were considered.
  • Population: studies involving students with special educational needs or disabilities across various educational levels (primary, secondary, and tertiary education).
  • Language: articles written in English or Spanish were included to ensure a comprehensive coverage of the relevant literature accessible to the research team.
  • Accessibility and Focus: studies that explicitly addressed AI’s role in enhancing accessibility, personalization, or equity within inclusive educational settings.
• Exclusion Criteria
  • Type of Studies: review articles, opinion pieces, editorials, unpublished theses, and conference abstracts were excluded to prioritize empirical evidence.
  • Temporal Relevance: studies published before 2021 were excluded due to the focus on recent developments.
  • Population: studies that did not involve students with disabilities or special educational needs, or those focusing solely on general education without an inclusive context.
  • Language: articles published in languages other than English or Spanish were not considered.

2.4. The Evaluation of the Quality of the Studies

The methodological quality of the studies was assessed using the Mixed Methods Appraisal Tool (MMAT), version 2018 (Hong et al., 2018), which allows for a consistent evaluation of qualitative, quantitative, and mixed-methods empirical studies within a single framework. The tool was selected due to its suitability for heterogeneous study designs and its structured application across the core quality criteria.

A total of seven full-text articles were subjected to quality appraisal. Each study was evaluated by two independent reviewers, and disagreements were resolved by a third reviewer. The evaluation considered five criteria specific to the methodological type of each study, such as the clarity of the research question, appropriateness of the design, bias control, and coherence in the reporting of results.

To tailor the evaluation to the objectives of this review, a sixth block of criteria was incorporated to assess the relevance and implementation of artificial intelligence (AI) in inclusive education. This block considered whether AI technologies were applied explicitly and actively to support inclusive practices, whether they enhanced accessibility or personalization, and whether sufficient technical detail was reported to assess their educational function.

It is important to note that these criteria were used exclusively for assessing the methodological quality and not as initial screening filters. However, only five of the seven studies met both the methodological and thematic standards required for final inclusion in the synthesis. The other two were excluded due to limitations in methodological transparency or the lack of an explicit connection between AI and inclusion.

• The inclusion criteria for quality appraisal were as follows:
  • Empirical studies published between 2021 and 2024;
  • Application of AI in inclusive education contexts involving students with disabilities or special needs;
  • Clear methodological reporting and presence of measurable educational outcomes.
• The exclusion criteria were as follows:
  • Theoretical articles or insufficient methodological detail;
  • Inconclusive or marginal use of AI in inclusive educational settings.

Table 2. MMAT evaluation of methodological quality and AI inclusion relevance.

Study (Author, Year)	Methodological Design	MMAT Criteria Met	Inclusion Status	Observations
Toyokawa et al. (2023)	Quantitative descriptive	5/5	Included	Strong methodological design with well-reported AI-enhanced reading tools. Clearly supports personalization and accessibility.
Tuna (2022)	Quantitative quasi-experimental	4/5	Included	AI use through humanoid robots is relevant and active, but randomization procedure was not clearly reported.
Sghaier et al. (2022)	Mixed methods	3/5	Included	Innovative AI–metaverse integration enhances access, but poor methodological integration between qualitative and quantitative strands.
Patiño-Toro et al. (2023)	Qualitative	5/5	Included	Inclusive MOOC design for the deaf community is clearly aligned with AI goals; triangulation of data sources could be improved.
Tzimiris et al. (2023)	Qualitative	5/5	Included	Rich contextual insight from parents; AI-supported remote education effectively addressed inclusion.
Starks and Reich (2023)	Qualitative	4/5	Excluded	Theoretical discussion of ethics and accessibility lacks empirical validation and does not apply AI within inclusive contexts.
Riley et al. (2024)	Qualitative	3/5	Excluded	Limited methodological detail and unclear link between AI and inclusive education.

summarizes the results of the MMAT evaluation, including the inclusion status of each study.

Several potential sources of bias were identified across the five studies included in the final synthesis. In the case of Tuna (2022), the main limitation was the lack of a clear description of the randomization process, which raises concerns regarding selection bias. The study by Sghaier et al. (2022), while innovative in its use of AI within a metaverse learning environment, presented weaknesses in the integration of its qualitative and quantitative components, limiting the interpretative coherence and increasing the risk of analytical bias. Patiño-Toro et al. (2023) lacked a robust triangulation of the data sources, potentially affecting the credibility and trustworthiness of its conclusions.

In contrast, Toyokawa et al. (2023) and Tzimiris et al. (2023) demonstrated a strong alignment between their research questions, design, and data interpretation, with a clear methodological transparency and contextual grounding, thus minimizing internal and external validity threats.

Additionally, a publication bias may be present, as most selected studies reported favorable outcomes with AI-based interventions. Some studies also lacked detailed contextual descriptions, which could limit the generalizability of the findings to other inclusive educational environments. However, no study was excluded solely due to a risk of bias; all decisions were based on the combined thematic alignment, methodological clarity, and relevance to the review objectives.

2.5. Article Selection Process

The initial search process yielded a total of 99 records, identified across four academic databases: Scopus (30), Web of Science (28), ERIC (21), and ScienceDirect (20). After removing 12 duplicates, 87 studies remained for title and abstract screening. At this stage, 80 articles were excluded because they did not meet the dual focus of this review. Specifically, many addressed only special educational needs without involving artificial intelligence (AI) or focused on AI in education without linking it to inclusion or students with disabilities.

To enhance transparency, the reasons for exclusion at this phase were categorized as follows:

(1)

A focus on AI without addressing inclusion or disabilities (29 articles), which centered on the general educational applications of AI without considering learners with special needs;

(2)

A focus on inclusion or special needs without AI integration (21 articles), which lacked any use or mention of artificial intelligence tools;

(3)

General educational studies not directly relevant to the review scope (20 articles), dealing with broader educational methods or tools unconnected to inclusive AI applications;

(4)

Non-empirical formats, such as reviews, opinion pieces, theses, or conference abstracts (10 articles), which did not meet the methodological inclusion criteria.

These exclusions ensured the thematic coherence and methodological rigor of the selected corpus, leading to a focused analysis aligned with the objectives of this review.

Seven studies proceeded to the full-text review and methodological assessment using the MMAT. Of these, five met all inclusion criteria and were retained for final analysis. The two excluded studies were rejected because they either lacked sufficient methodological detail or did not demonstrate an explicit and active application of AI in inclusive education.

The selection process is summarized in the PRISMA 2020 flowchart (Figure 7), which is structured according to the standard phases of identification, screening, eligibility, and inclusion. The search and screening workflow was supported by Parsifal, a systematic review management platform that facilitated the integration of search results, duplicate removal, and reviewer collaboration.

Although the initial search retrieved 99 potential studies, only 5 met all the established criteria for inclusion. This limited number is not due to a lack of relevant literature, but rather the result of rigorous inclusion criteria focused on empirical studies that specifically address artificial intelligence applications in inclusive education settings. Most excluded studies either lacked methodological robustness, focused on general education without a reference to inclusion, or addressed AI applications unrelated to students with disabilities or special needs. This outcome reflects both the novelty of the topic and the current scarcity of high-quality empirical research that integrates AI with inclusive education practices.

2.6. Characteristics of the Selected Studies

To provide a more detailed overview of the included studies,

Table 3. Main characteristics of the selected articles.

Lead Author (Year)	Country	Type of AI Intervention	Main Findings
Toyokawa et al. (2023)	Japan	AI applied to active and analytical reading	AI helps identify stumbling points and strengths in student learning.
Tuna (2022)	Turkey	Use of humanoid robots and virtual agents	Humanoid robots enhance symbolic play skills in children with autism.
Sghaier et al. (2022)	Saudi Arabia	Intelligent system based on the metaverse	The metaverse improves immersion and performance for students with disabilities.
Patiño-Toro et al. (2023)	Colombia	MOOCs designed for the deaf community	Well-designed MOOCs facilitate inclusive access to education for deaf individuals.
Tzimiris et al. (2023)	Greece	Emergency remote teaching supported by AI	Remote teaching had a negative impact on students with special needs.

presents the main characteristics of each article, including the author, year of publication, country, type of AI intervention, and the main findings in terms of accessibility, the personalization of learning, and challenges associated with the implementation of AI in inclusive education.

2.7. Data Analysis

A qualitative content analysis was performed using ATLAS.ti, a widely adopted software for organizing and analyzing textual data in research. The research questions guided the development of thematic categories such as “accessibility”, “personalization of learning”, “ethical challenges”, and “resistance to technology adoption”.

For instance, under the category of Accessibility, passages describing screen readers or adaptive systems were coded. In Personalization, excerpts related to intelligent tutoring systems were included for their ability to adjust the content and pace. Similarly, Ethical Challenges included concerns around data privacy and algorithmic bias.

The analysis was carried out in multiple phases. First, all documents were imported into ATLAS.ti and reviewed line by line to extract the relevant content. These excerpts were initially assigned to predefined categories, which were refined iteratively as new themes emerged. For example, within Accessibility, subthemes such as technological tools and infrastructure limitations were identified.

To ensure analytical reliability, the coding process was independently reviewed by two researchers. This double-coding approach reinforced the credibility of the thematic framework. In cases where discrepancies emerged, external AI-based language models were used as complementary tools to support the triangulation of coder interpretations, helping to refine the definition of categories and ensure consistency across the dataset. The integration of this auxiliary AI input contributed to a more transparent and balanced interpretation of the findings, while preserving the critical judgment of the research team.

To complement the qualitative analysis, a tree map was generated based on word frequency (threshold: ≥40 occurrences), using base word forms to maintain linguistic consistency. This visualization highlighted key terms such as use, learn, and design, which clustered under broader themes like accessibility and personalization, aligning with the study’s central questions.

This combination of researcher-driven coding, AI-supported validation, and quantitative visualization provided a robust framework for exploring the dataset. The tree map served as a bridge between the qualitative insights and broader patterns in the literature, enhancing our understanding of how AI tools are being applied to advance inclusive education.

3. Results

The results of this systematic review are organized according to the guiding research questions, focusing on how artificial intelligence (AI) supports accessibility and personalization in inclusive education and identifying the challenges faced by institutions in its implementation. Each set of findings is accompanied by figures that illustrate key trends and patterns across the literature.

RQ1: How does artificial intelligence enhance accessibility and the personalization of learning for students with special needs or disabilities?

The studies reviewed concur that AI facilitates both accessibility and personalization in inclusive learning environments. However, each study highlights distinct ways in which these technologies contribute to more equitable and adaptive educational practices for students with disabilities. For instance, Toyokawa et al. (2023) and Patiño-Toro et al. (2023) emphasize the use of adaptive educational platforms to foster inclusion.

Toyokawa et al. (2023) show that AI enhances active reading by identifying students’ learning strengths and challenges in real time, enabling educators to tailor content and strategies to individual needs. Meanwhile, Patiño-Toro et al. (2023) demonstrate that MOOCs designed for deaf students can eliminate communication barriers and expand equitable access to higher education.

Tuna (2022) provides a complementary perspective by exploring the use of humanoid robots and virtual agents in the education of children with Autism Spectrum Disorder (ASD). The findings indicate that these technologies not only personalize learning by adapting to the pace and style of the student but also foster cognitive and social skills, creating a meaningful environment that addresses both their academic and emotional needs. Furthermore, technologies such as virtual agents allow students with disabilities to interact in simulated social contexts, developing communication skills and enhancing their confidence in real-life situations. Together, these applications demonstrate AI’s potential to create inclusive and meaningful educational experiences in various learning contexts.

Despite these benefits, several studies note that the effectiveness of AI depends on infrastructure availability and teacher preparedness. Without adequate resources and professional development, the potential of AI to improve inclusion remains limited. High costs and technological complexity may also hinder its implementation in low-resource contexts, underscoring the need for context-sensitive approaches to ensure equitable access.

Most studies focus on how AI enhances accessibility and the personalization of learning. However, Sghaier et al. (2022) explore the metaverse as an immersive educational environment, emphasizing its ability to offer interactions tailored to individual needs, especially for students with motor or communication disabilities. Although its implementation faces technological challenges, these immersive environments, along with intelligent tutoring systems, enable the teaching to be adjusted to each student’s pace and progress. These applications highlight AI’s potential to create inclusive and meaningful educational experiences in diverse learning contexts.

On the other hand, Tzimiris et al. (2023) adopt a contextual approach, analyzing how remote teaching and digital platforms facilitated or complicated educational access for students with special needs during the pandemic. Although this study does not address the personalization of learning through AI, it underscores that a continuous connection to educational environments was crucial despite the difficulties stemming from a lack of technological resources and parental resistance to using digital tools.

These results collectively indicate that while AI has the potential to transform inclusive education, its effectiveness is influenced by broader systemic factors, such as digital inequality and sociocultural acceptance. Addressing these barriers will require collaborative efforts among educators, policymakers, and technology developers to design solutions that are both innovative and accessible.

Figure 8 provides an integrated view of the analyzed studies, highlighting how AI and emerging technologies transform educational models by adapting to students’ individual needs. The key elements identified include MOOCs, which eliminate barriers for students with specific needs, and the diversity of educational materials, which facilitates active participation in learning. Furthermore, flexible pedagogical strategies that enable pressure-free learning, adjusted to each student’s pace, are emphasized.

Additionally, Figure 8 incorporates the role of the metaverse as an immersive environment enriching the educational experience, highlighting the need for adequate infrastructure and technological acceptance to ensure effective inclusion. Together, the figure synthesizes how the reviewed studies agree that AI and these technologies remove barriers, facilitate access, and promote accessible, adaptive, and student-centered learning environments.

RQ2: How does AI help reduce teachers’ administrative workload and improve their interaction with students in inclusive environments?

According to the reviewed studies, 60% (3/5) of the authors highlight how AI alleviates teachers’ workload by automating repetitive tasks, such as grade management and academic progress tracking. This enables educators to dedicate more time to meaningful pedagogical activities, enhancing the learning experience and improving job satisfaction. Toyokawa et al. (2023) point out that automating the monitoring of academic performance facilitates more timely and personalized interventions, allowing teachers to focus on direct student engagement.

Tuna (2022) emphasizes that humanoid robots and virtual agents not only support personalized learning for students with Autism Spectrum Disorder (ASD) but also assist teachers in planning activities, reducing their administrative burden and enabling better pedagogical support.

In remote education contexts, Tzimiris et al. (2023) highlight that digital platforms helped maintain educational continuity during the pandemic. However, they also identified challenges, such as parents’ lack of familiarity with technology, which in some cases affected student participation. Additionally, data literacy, promoted among both teachers and students, is key to preparing them for a future where decision-making based on data will be essential.

Despite these benefits, the increasing reliance on AI for administrative tasks raises concerns about over-automation, which could reduce opportunities for direct human interaction in educational settings. These findings highlight the need to strike a balance, ensuring that AI complements rather than replaces pedagogical engagement. Moreover, successful implementation still depends on the investment in infrastructure and teacher training, especially in low-resource contexts.

Although not all studies explicitly focus on the administrative workload, Patiño-Toro et al. (2023) concentrate on inclusion through the creation of accessible MOOCs. This approach, more oriented towards inclusive educational design, reflects how AI can facilitate the participation of students with specific needs, even if it does not directly address administrative optimization.

Sghaier et al. (2022) point out that the metaverse allows teachers to efficiently manage personal data and coordinate activities, streamlining information transfer and task planning. This immersive technology not only optimizes resource management but also enhances learning environments by providing more dynamic educational experiences.

Figure 9 synthesizes how AI acts as a key factor in reducing the administrative workload and promoting social and professional impact in inclusive educational environments. It highlights that task automation allows teachers to spend more time on pedagogical interaction and individualized support while promoting adaptive learning environments and data literacy to prepare both students and teachers for the challenges of modern education.

The figure also illustrates how AI facilitates efficient resource and time management, educational and career guidance, and the transfer of personal data. However, it also underscores the importance of consent and data literacy for ethical and effective use, emphasizing that the impact of these technologies requires collaboration among teachers, students, and families to maximize their potential.

RQ3: What are the main challenges and limitations educational institutions face when implementing AI in inclusive environments?

The implementation of AI in inclusive education faces technological, pedagogical, and social challenges that limit its adoption and sustainability. Figure 10, Figure 11 and Figure 12 illustrate how these challenges relate to individual and family needs, as well as to digital and technological learning processes, highlighting the complexities institutions must address to ensure effective inclusion.

The first set of challenges focuses on the individual needs of students and family limitations. The lack of digital accessibility, as noted by Patiño-Toro et al. (2023), can act as a barrier, particularly for students with hearing impairments who depend on specialized resources such as accessible MOOCs. Additionally, Tzimiris et al. (2023) emphasize parental resistance to the use of digital technologies, which can limit student participation in remote education.

Figure 10 suggests that fostering parental involvement and providing continuous support are essential, along with offering accurate educational assessments to identify and address the specific needs of each student. Another significant obstacle is the lack of devices and connectivity, which restricts participation for students in vulnerable situations. This challenge is particularly critical in remote teaching contexts, where the reliance on technological tools can deepen educational inequalities. Similarly, identifying and monitoring difficulties is crucial for providing timely support and avoiding the exclusion of students with special needs.

Moreover, these findings highlight the importance of designing inclusive policies that address not only technological barriers but also the sociocultural factors influencing the adoption of AI in education. Developing locally adapted solutions and fostering collaboration among key stakeholders can mitigate these challenges, ensuring that AI benefits all students equitably.

The second set of challenges focuses on the impact of emerging technologies on digital learning processes. Sghaier et al. (2022) emphasize that implementing the metaverse in inclusive education requires adequate technological infrastructure and specialized teacher training. The figure suggests that data literacy and the ability to ethically manage and use information are essential skills for both teachers and students, preparing them for an environment based on informed decision-making.

Tuna (2022) addresses the challenges of using humanoid robots and avatars, which can enhance cognitive and social skills in students with Autism Spectrum Disorder (ASD). However, their integration into the classroom is limited by high costs and technical complexities. Additionally, Figure 12 highlights the importance of tools like LEAF and learning management systems, which facilitate personalized feedback and the adaptation of content to meet each student’s needs.

On the other hand, Tzimiris et al. (2023) stress how ethical dilemmas related to the use of personal data complicate the implementation of technologies in inclusive education. Establishing clear policies on consent and privacy is essential to ensure that emerging technologies are used safely and equitably.

Figure 12 presents the most frequently occurring terms in the analyzed articles, highlighting key patterns in the research on the application of AI to inclusive education. The prominence of words such as “use” (200), “learn” (197), “base” (93), “design” (77), “develop” (70), and “improve” (53) underscores a pragmatic focus, indicating how AI is employed to optimize educational processes that foster inclusive learning.

Similarly, terms like “provide” (69), “include” (67), and “make” (45) reflect the potential of AI to create accessible and balanced learning environments, ensuring that students have the necessary resources to fully participate. Meanwhile, words such as “read” (48), “show” (48), “create” (44), “consider” (44), and “take” (42) point to the design of innovative learning experiences tailored to individual needs.

Additionally, the relevance of “special”, “personal”, “educational”, and “help” emphasizes a learner-centered perspective, aiming to personalize instruction and support social interaction according to the unique characteristics of each student. Overall, the frequency of these terms underscores the potential of AI to promote active student engagement, while establishing inclusive educational contexts that adapt to the evolving demands of contemporary teaching and learning.

The findings of the study demonstrate that AI enhances the accessibility and personalization of learning, fostering the active engagement of students through MOOCs and adaptive technologies. Furthermore, it alleviates the administrative burden on educators by automating tasks, thereby enabling them to prioritize individualized pedagogical support and enhance their data literacy. However, significant obstacles remain, including the absence of specialized teacher training, parental reluctance to embrace technology, and the substantial costs and infrastructure requirements associated with advanced tools such as robots and the metaverse.

Table 4. Main findings by research question.

References	Research Question	Finding
(Toyokawa et al., 2023; Sghaier et al., 2022)	RQ1	AI facilitates active and personalized participation through MOOCs and adaptive technologies, improving accessibility, autonomy, and the educational experience for students with disabilities.
(Tzimiris et al., 2023)	RQ2	AI significantly reduces the administrative workload by automating repetitive tasks, allowing teachers to focus on individualized support. It also promotes data literacy for both students and teachers.
(Tuna, 2022; Patiño-Toro et al., 2023)	RQ3	The main challenges include the lack of inclusive equipment, parental resistance to technology use, and the need for specialized teacher training. However, AI enhances accessibility in MOOCs and improves cognitive and social skills for students with ASD through interaction with robots.

provides a summary of these findings, linking the responses to each research question with the main aspects identified in the analyzed studies.

4. Discussion

This systematic review examined the role of artificial intelligence (AI) in transforming inclusive education. The findings show that AI enables the personalization and adaptation of learning environments to meet students’ individual needs—an essential component of inclusive education (Toyokawa et al., 2023; Patiño-Toro et al., 2023). Tools such as MOOCs and adaptive platforms help remove communication barriers and promote equitable access for students with disabilities (Woolf et al., 2013; Toyokawa et al., 2023; Sein-Echaluce et al., 2016). These results align with previous research that highlights AI’s contribution to accessibility through educational tools that account for diverse learning needs (Gródek-Szostak et al., 2023; Habib et al., 2023; Osuna Acedo & Tejera Osuna, 2016).

However, while these benefits are promising, they are contingent upon addressing systemic challenges, such as digital inequality, a limited access to infrastructure, and the availability of trained personnel. These factors are critical for ensuring that the potential of AI is realized equitably across diverse educational contexts.

For example, Lin and Chang (2024) emphasize that access to technological infrastructure and teacher training are essential prerequisites for maximizing the benefits of AI. In contrast, our findings place less emphasis on these contextual barriers, suggesting that perceptions and priorities may vary depending on the educational setting. This divergence highlights the importance of tailoring implementation strategies to the specific realities of each context.

This review also shows that AI reduces teachers’ administrative workload, enabling them to dedicate more time to personalized pedagogical support (Zahurin et al., 2024; Patiño-Toro et al., 2023). While previous studies—such as Zahurin et al. (2024) and Tzimiris et al. (2023)—emphasize automation of routine tasks, our findings suggest that such efficiencies also allow for more responsive, student-centered instruction. However, care must be taken to avoid an excessive reliance on automation, which may limit opportunities for meaningful teacher–student interaction. AI should complement, not replace, human engagement.

While Tuna (2022) focuses on the benefits of humanoid robots and virtual agents in supporting students with Autism Spectrum Disorder (ASD), our findings expand this view by demonstrating their value for a broader range of students. These tools not only enhance academic learning but also contribute to socio-emotional development—an essential component of inclusive education (Knox et al., 2019). However, barriers such as parents’ lack of technological skills and resistance to digital tools continue to limit the effectiveness of these innovations, particularly in remote contexts (Monova-Zheleva & Prodanov, 2024). These findings underscore the importance of engaging families through training and support to ensure equitable participation (Shourbagi, 2017; Soodak & Erwin, 2000; Alyokhina & Shemanov, 2023).

Regarding implementation challenges, significant technological, pedagogical, and social limitations were observed in adopting AI in inclusive settings. One major challenge is the lack of adequate infrastructure and specialized training, especially in institutions with limited resources (Kasinathan, 2020; Salas-Pilco et al., 2022). Consistent with these results, previous studies have highlighted that while AI and emerging technologies like the metaverse offer new opportunities for immersive learning, their adoption depends on a robust technological infrastructure and continuous teacher training (Sghaier et al., 2022; Schultz et al., 2016). Without substantial investment in these areas, the scalability and long-term sustainability of these technologies remain limited, particularly in under-resourced regions.

The success of AI in inclusive educational environments depends on continuous teacher training in digital competencies and AI literacy. Specific training in AI will enable educators to use these tools ethically and effectively, maximizing their pedagogical benefits in inclusive education (Gibellini et al., 2023). Investment in professional development ensures sustainable implementation and optimizes the use of these technologies in classrooms, promoting more inclusive learning. Additionally, the development of low-cost and scalable training programs could mitigate barriers in underserved regions, fostering equity in the access to these technologies.

Without accessible technological infrastructure, the inclusion of students with special needs can be compromised and the benefits of AI diminish (Sakharova et al., 2024). This issue is particularly relevant in under-resourced geographic areas, where limited access to technological devices and poor connectivity restrict equitable student participation (Gonzales et al., 2020; Alam & Forhad, 2023).

The adoption of advanced technologies, including humanoid robots, remains limited due to the significant expenses associated with their acquisition and maintenance (Monova-Zheleva & Prodanov, 2024). This observation aligns with other studies that emphasize that a lack of training in using advanced tools may prevent teachers from fully leveraging AI’s potential to facilitate inclusive learning (Gibellini et al., 2023; Lin & Chang, 2024; Irdamurni et al., 2019; Vanderpuye et al., 2020). To address these challenges, the establishment of public–private partnerships and the promotion of affordable technological solutions, such as open-source software, could help bridge these gaps and ensure that AI contributes effectively to inclusive education.

From an ethical standpoint, the use of AI in inclusive education requires clear policies to safeguard data privacy and student rights. The collection of personal information, especially involving vulnerable populations, raises serious concerns (Mandinach & Jimerson, 2021). Ensuring informed consent, transparent data use, and algorithmic fairness is essential (Sghaier et al., 2022; Klimova et al., 2023; Tzimiris et al., 2023; Mandinach & Jimerson, 2021; Devi et al., 2023). Training in data literacy can empower educators and students to make responsible decisions and mitigate bias.

Moreover, addressing algorithmic bias is critical to prevent unintended discrimination against students with disabilities. Ensuring transparency in AI development and conducting regular audits can mitigate these risks and enhance the trustworthiness of these tools.

Finally, the findings of this study highlight that collaboration among teachers, students, and families is a key factor in maximizing AI’s positive impact on inclusive education. Unlike previous studies that emphasize parental support in isolation (Tzimiris et al., 2023; Shourbagi, 2017) our results suggest that the integral collaboration of the entire educational community (teachers, students, and families) is essential for technology acceptance and the continuity of learning in AI-assisted environments. Furthermore, while studies like those by Vanderpuye et al. (2020) and Lakkala et al. (2021) point to the general benefits of digital platforms, this study emphasizes that the effectiveness of these tools increases significantly when applied in combination with strategies that promote active participation across the educational community and the development of digital skills among all its members. According to our findings, this integration is fundamental for creating an inclusive, ethical, and sustainable approach that enables effective and equitable AI implementation in education.

In this regard, to directly address the challenges identified, several specific and practical strategies are suggested:

• Establishing public–private partnerships to fund the infrastructure and provide specialized training for teachers in low-resource regions;
• Implementing continuous training programs in digital literacy and AI for both teachers and families, promoting ethical data use and technology acceptance;
• Creating communities of practice and collaborative networks that facilitate the exchange of experiences and best practices among educational institutions in different contexts;
• Adopting low-cost technological solutions (such as open-source software or simpler devices) to minimize access gaps and ensure sustainability.

Furthermore, fostering interdisciplinary research and collaboration between educators, policymakers, and technology developers is essential to create context-sensitive solutions that address the unique challenges of diverse educational settings.

5. Limitations and Future Lines of Research

It is important to acknowledge the limitations of the present study in order to contextualize the findings appropriately. First, the scope of this study was limited to articles published in specific databases, including Scopus, Web of Science, and ERIC, between 2021 and 2024. Although these sources are esteemed for their quality, it is conceivable that some pertinent works published in alternative repositories or beyond the specified timeframe were not incorporated. This exclusion may have resulted in a narrow representation of the available evidence, particularly from non-English language studies and gray literature that could provide additional insights into regional or localized applications of AI in inclusive education. Furthermore, the qualitative approach based on a literature review lacks supplementary empirical studies, which limits the ability to directly measure the impact of artificial intelligence in real educational settings. The contextualization of the results is similarly constrained to specific educational systems, and any extrapolation to other contexts should be approached with caution. Additionally, the relatively emergent nature of this topic restricted the volume of the research available, underscoring the need for further scholarly inquiry to expand the evidence base in this field.

Another significant limitation is the potential for publication bias, as studies reporting positive results tend to be more visible in the scientific literature, which may limit the analysis of negative experiences or challenges associated with AI implementation in inclusive education. Furthermore, although this review identified the resistance from teachers and families to adopting these technologies, deeper empirical investigations are necessary to analyze the underlying factors of this resistance and the barriers institutions face in effectively incorporating AI. For instance, the sociocultural dynamics influencing the acceptance of AI tools in different regions remain underexplored and warrant further investigation to design culturally sensitive implementation strategies.

Mitigating this publication bias in future research involves broadening the scope of the consulted sources by incorporating additional databases or institutional repositories that house gray literature, as well as conducting searches in multiple languages to capture studies not available in English. Likewise, designing review protocols with clear quality assessment criteria and well-defined inclusion and exclusion parameters allows for a more balanced perspective on both positive and negative outcomes in the study of AI within inclusive contexts. Such protocols should also consider interdisciplinary approaches that combine educational, technological, and sociological perspectives to address the multifaceted nature of AI integration in education.

These limitations give rise to potential avenues for future research. First and foremost, it is imperative to conduct empirical studies within authentic educational settings to assess the efficacy of AI in enhancing learning and inclusion. Furthermore, the creation of novel AI-based pedagogical tools that facilitate accessibility and the personalization of learning is imperative. These tools should prioritize adaptability to diverse educational contexts, particularly in regions with limited resources, to ensure equitable access to the benefits of AI. A longitudinal investigation of the impact of these technologies on students’ academic and socio-emotional development would be beneficial in order to gain a deeper understanding of the ways in which AI influences the holistic progress of students with diverse needs.

Finally, the implementation of AI should be explored through a comparative perspective across different cultures and educational contexts, with a particular focus on equity and ethics in the use of these technologies. This approach will guarantee the long-term viability of AI integration in inclusive education, ensuring that it contributes to the overall well-being of students in an inclusive manner. Such comparisons can also illuminate best practices that can be adapted and scaled to various educational systems, promoting a global understanding of effective AI integration.

Moreover, it is imperative to enhance the teacher training in AI competencies and data literacy, thereby equipping educators with the knowledge and skills to utilize these tools ethically and effectively in their pedagogical practices. The reinforcement of such professional development will optimize the potential of AI technologies, thereby fostering inclusive and impactful educational experiences. This training should also include modules on ethical considerations and the management of sensitive student data, ensuring that educators are prepared to navigate the challenges associated with AI use in education.

Additionally, exploring tailored implementation strategies in regions with limited resources could further illuminate cost-effective and context-specific solutions. This may involve pilot programs that assess the feasibility of low-cost technologies, continuous teacher training models, and partnership building with local communities to ensure that the AI solutions align with the unique sociocultural dynamics of each setting. Collaborations between local governments, non-profit organizations, and technology developers could help bridge gaps in the infrastructure and capacity, fostering sustainable AI adoption in inclusive education.

6. Conclusions

This systematic review confirms that artificial intelligence offers meaningful contributions to inclusive education by improving accessibility, supporting personalized learning, and reducing pedagogical barriers for students with disabilities. Tools such as intelligent tutoring systems, adaptive MOOCs, virtual agents, and immersive environments demonstrate the capacity to tailor instruction, remove communication obstacles, and promote autonomy in learning.

Beyond the benefits for students, AI also alleviates teachers’ administrative workload, allowing them to concentrate on high-impact pedagogical strategies. The automation of data tracking and lesson planning contributes to more efficient educational environments and enhances the quality of teacher–student interactions. These changes support a shift toward more responsive and equitable teaching practices.

Nonetheless, the integration of AI in inclusive contexts remains challenged by structural and ethical limitations. The limited access to infrastructure, lack of teacher training in digital tools, and family resistance to technological interventions can hinder effective implementation. Moreover, concerns about data privacy and algorithmic bias demand the development of clear governance frameworks that prioritize ethical and equitable use.

In response to these findings, there is a need for inclusive strategies that combine accessible technological solutions, professional development, and the active involvement of all stakeholders. Future research should focus on generating robust empirical evidence, especially in underserved regions, and promoting interdisciplinary collaboration that ensures AI technologies genuinely support inclusive and sustainable education.