Abstract
Artificial Intelligence (AI) has rapidly entered higher education, transforming how students learn, communicate, and access knowledge. From essay generation to virtual tutoring, AI now mediates much of the contemporary academic experience. Proponents argue that these technologies can democratise education by bridging long-standing achievement gaps among students from low-income, rural, or first-generation backgrounds. Nevertheless, despite these claims, there is limited empirical understanding of how marginalised students actually experience and engage with AI in their learning contexts. This study addresses this gap by examining how marginalised students at the University of Johannesburg (UJ) utilise AI tools such as ChatGPT, Copilot, Grammarly, Bing Chat, Google Bard, Anthropic’s Claude, AlphaCode, DALL·E 2, and Midjourney alongside digital platforms like Microsoft Teams, Moodle, and Zoom for academic purposes.
Using a qualitative research design, semi-structured in-person and online interviews were conducted, transcribed, and thematically analysed. The findings reveal that while AI tools offer personalised learning opportunities and can enhance academic performance, they do not overcome entrenched structural inequalities, including digital poverty, low digital literacy, and inadequate institutional support. Consequently, AI integration without parallel systemic interventions risks reproducing existing educational hierarchies in more subtle ways. The study concludes that the transformative promise of AI in education will remain unrealised unless it is grounded in principles of inclusion, equity, and social justice. This research contributes to critical debates on AI and educational equity in South Africa, offering insights into how technology can either reinforce or reimagine existing structures of advantage and exclusion.