The intersection between academic and AI literacy : the role of scaffolded support in developing first-year student agency in writing

Insufficient academic literacy is a significant predictor of academic success and therefore requires targeted support within higher education. However, the emergence of Generative AI (GenAI) has complicated the identification of students' support needs, as well-formulated AI-generated texts can mask deficiencies in critical thinking and academic writing skills. To date, limited empirical research has examined how AI literacy and academic literacy intersect within structured support programmes, particularly in the South African and Afrikaans-speaking context. This study investigates how an AI-supported approach can enhance the academic literacy, writing proficiency, and critical agency of first-year students. Based on the research findings, a conceptual framework is proposed in which academic literacy forms the foundation, AI literacy functions as a mediating competency, and AI-mediated writing practices constitute the interaction layer in which human judgement and responsibility remain central. Situated within the pragmatic paradigm, a phased mixed-methods research design was employed to evaluate the Write It+ academic literacy support programme across five implementation cycles while simultaneously investigating an AI-supported writing intervention. Within this intervention, students produced successive versions of an academic paragraph using AI feedback aligned with the assessment rubric. Data collection included pre- and post-tests using the Toets van Akademiesegeletterheidsvlakke (TAG), analyses of successive text versions, AI interaction data, and a questionnaire measuring self-efficacy, processing fluency, and critical agency. The findings demonstrate that the Write It+ programme leads to significant improvements in the academic literacy of first-year students and substantially reduces the proportion of students classified as high risk. In addition, the quality of students' academic writing improved significantly across successive text versions. Although GenAI provided targeted support throughout the writing process, students remained active authors of their texts by critically evaluating, adapting, or rejecting AI-generated suggestions. Furthermore, a moderate positive relationship was found between self-efficacy and critical agency, suggesting that structured interaction with GenAI not only strengthens students' confidence but also enhances their ability to evaluate AI-generated output critically. This dissertation makes three important contributions to applied linguistics and digital pedagogy. First, it positions AI literacy as an integrated subdomain of academic literacy within higher education. Second, it provides empirical evidence for the effectiveness of the Write It+ programme at scale, an aspect that has not previously been investigated systematically. Third, it introduces the HAIL model (Human-centred AI-mediated Academic Literacy), a human-centred conceptual model in which AI is employed to support academic literacy without displacing human agency from the writing process. In doing so, this study advocates a process-oriented approach to academic writing in which the student, rather than the technology, retains ultimate responsibility for the final text.