Artificial Intelligence Chatbots for Inclusive Pedagogical Practices: A PRISMA Systematic Review and Bibliometric Mapping (2021–2025)

Abstract

The integration of artificial intelligence (AI) in education has accelerated, yet evidence on how AI chatbots support inclusive pedagogical practices remains fragmented. This study synthesizes recent literature on the contributions of AI chatbots to inclusive education, emphasizing accessibility, equity, and pedagogical adaptation. We conducted a systematic review guided by PRISMA 2020. A sensitive search in November 2025 retrieved 338 records from Scopus (n = 59) and Dimensions (n = 279). After removing duplicates (n = 13) and applying eligibility criteria (2021–2025; open access; relevance to inclusive education and AI chatbots), 11 studies were included for in-depth analysis. In parallel, bibliometric mapping was conducted using VOSviewer to describe publication trends, keyword co-occurrence, collaboration networks, and citation/co-citation structures. Findings suggest three contributions of AI chatbots to inclusive pedagogical practices: (i) personalization and instructional adaptation, where chatbots operate as intelligent tutors and scaffold learning pathways for diverse learners; (ii) functional accessibility, including speech-to-text and text-to-speech affordances that reduce participation barriers for students with sensory or communication needs; and (iii) equity-oriented support, in low-resource contexts where chatbots may partially compensate for limited instructional capacity. However, the evidence base is constrained by small samples, short-term implementations, and heterogeneous outcomes, limiting generalizability. A persistent gap concerns support for students with reduced mobility, alongside ethical risks related to privacy, academic integrity, dependency, and potentially perpetuating medicalized views of disability. We conclude that AI chatbots can strengthen inclusive education when implemented under principles of equity, transparency, and pedagogical ethics and when aligned with Universal Design for Learning and reasonable accommodation (reasonable adjustments).