The rapid rise of generative Artificial Intelligence (AI) such as ChatGPT and DeepSeek has brought both excitement and uncertainty into educational spaces worldwide (Zawacki-Ritcher et. al., 2019). In Singapore, and particularly within the Dyslexia Association of Singapore (DAS), educators are increasingly reflecting on how AI can be used meaningfully to support teaching and learning without compromising foundational skills, student’s independence, or academic integrity. Given that many DAS students experience learning difficulties, careful consideration is needed to ensure that AI functions as a scaffold rather than a shortcut.

This article explores the use of AI for lesson planning by educators and the use of AI for students’ assignments, drawing on Ng, E. N.’s (2024) paper on AI-Enhanced Lesson Design Process and the systematic review by Pande, K. and Teng, K. E. (2024) on generative AI in self-directed learning. The discussion is grounded in DAS pedagogical principles, including explicit instruction, structured scaffolding, metacognitive development, and inclusive education. More recently, educators and researchers have also begun exploring Retrieval-Augmented Generation (RAG), a technique that combines document retrieval with generative models to improve factual accuracy and reduce hallucinations in AI outputs (Lewis et al., 2020; Xie et al., 2025; Li & Wong, 2025).​ RAG, in comparison to conventional generative AI, is like a student who consults credible reference materials before writing an essay, ensuring accuracy and depth, whereas conventional generative AI resembles a student relying solely on memory. This retrieval step allows RAG to generate responses that are not only contextually rich but also verifiable and grounded in reliable information.

AI for Lesson Planning in DAS Classrooms

Lesson planning at the DAS is a deliberate and reflective process. Educators design lessons that are highly structured and responsive to different learners’ profiles. This often involves breaking down concepts into smaller steps and planning multiple layers of scaffolding. Generative AI can assist in this process by rapidly producing lesson outlines aligned to learning objectives, suggesting differentiated tasks and generating practice questions.

Ng, E. N. (2024) provides a useful framework for integrating AI into lesson planning while preserving professional judgement. According to Ng, E. N. (2024), AI should be embedded within a cyclical design process rather than be used as an autonomous planner. Educators begin by crafting contextualised prompts that include information about students’ needs, prior knowledge and instructional goals. AI then generates an initial lesson draft, which is critically evaluated by educators using pedagogical criteria before being refined through professional judgement. As seen in Figure 1, when RAG is used in this process, the AI system first retrieves relevant curriculum documents, structured literacy resources, or DAS-produced materials before generating lesson ideas, making it easier for teachers to keep activities aligned with existing programmes and evidence-based practices (Lewis et al., 2020).​

Figure 1. A diagram to demonstrate the role of RAG and generative AI for educators.

This approach aligns strongly with the DAS practice. Like ChatGPT, the DAS Main Literacy Programme (MLP) Curriculum Crafter (MCC) is an AI-based tool designed to assist educators with planning and content generation. AI may suggest activities or explanations, but it cannot fully account for the working memory limitations, phonological processing difficulties, or emotional experiences of students with dyslexia (Ng, E. N., 2024). Educator’s mediation is therefore essential to ensure that lessons remain accessible, explicit, and purposeful.

Despite its benefits, AI also presents its limitations. AI-generated lessons may be overly generic or misaligned with structured literacy approaches. There is also a risk that over-reliance on AI could reduce educators’ engagement in reflective lesson design. Within the DAS, AI should therefore be positioned as a planning assistant that enhances efficiency and not a pedagogical authority. RAG does not remove these concerns entirely, but by grounding generation in a defined repository (e.g. syllabi, workbooks and handouts), it can make outputs more transparent and easier for teachers to verify against trusted sources (Huang et al., 2024).​

AI for Student Assignments and Learning

From the student’s perspective, generative AI offers immediate access to explanations, examples, and feedback. Pande, K. and Teng, K. E.’s (2024) systematic review on generative AI in self-directed learning highlights that AI tools can support learners in key processes such as goal setting, planning, monitoring understanding, and reflection. These processes are particularly relevant for DAS students, who may struggle with task initiation, organisation, and comprehension.

When guided appropriately, AI can support students in breaking down assignment instructions, clarifying vocabulary, generating ideas, and reviewing drafts. For example, a student may use AI to ask for a simplified explanation of a concept, to brainstorm possible story ideas, or to receive feedback on sentence clarity. These uses align with DAS’s emphasis on metacognition and self-awareness, helping students to reflect on their learning rather than simply producing an end product.

In a RAG-based setup, student queries are answered using information drawn from a curated educational knowledge base, such as school-approved notes or readings before the AI generates explanations, which can lower the risk of misleading or fabricated responses and make it easier for teachers to trace where information came from (International Journal of Teaching, Learning and Education, 2024; Huang et al., 2024).​

However, the use of AI in assignments also raises significant concerns. One key risk is reduced cognitive engagement. If students rely heavily on AI to generate answers or complete tasks, they may bypass productive struggle, which is essential for building understanding and resilience. There is also the risk of superficial performance, where AI-generated responses appear fluent but mask weak conceptual understanding.

Furthermore, generative AI can produce confident yet incorrect explanations or fabricated information, which may be difficult for students with learning difficulties to detect. Finally, issues of academic integrity arise when students submit AI-generated work as their own, undermining the purpose of assessment and limiting opportunities for meaningful feedback. Even with RAG, retrieval quality and source selection remain critical; poor or outdated repositories can still lead to inaccurate or unhelpful explanations, so teacher oversight and explicit instruction in critical reading are still required (International Journal of Teaching, Learning and Education, 2024).​

Reframing AI Use at DAS

To harness the benefits of AI while mitigating risks, it is important to establish clear expectations around its use. A guiding principle at DAS is that AI should support thinking, not replace it. Appropriate uses of AI include asking for clarification of instructions, requesting feedback on a student’s own draft, generating additional practice questions, and identifying areas for improvement.

In contrast, high-risk uses include asking AI to complete assignments for submission, copying responses with minimal personal input, and relying on AI-generated references without verification. Explicitly teaching students how to use AI ethically and reflectively is therefore an important component of digital literacy. This can include helping students recognise when an answer is grounded in retrieved sources (as in RAG) and when it is purely generative, and to cross-check both against teacher-provided materials.

Integrating AI Thoughtfully into Teaching and Learning

Generative AI is becoming increasingly prevalent worldwide, with rapid growth in its adoption across both educational contexts and the workplace (Zawacki-Ritcher, et. al., 2019; Kasneci, et. al., 2023). Advances in large language models have accelerated the integration of AI into everyday academic and professional practices, influencing how information is accessed, processed, and communicated. As these technologies become more embedded in global work and learning environments, educators are increasingly required to consider how students can be guided to engage with AI in ways that support meaningful learning, sustained cognitive engagement, and responsible use, particularly for learners who benefit from explicit instruction, structured support, and the development of metacognitive awareness. RAG is one promising direction in this landscape, as it aims to make AI systems more accurate, updateable and accountable by linking responses back to external sources (Lewis et al., 2020; Li & Wong, 2025).​

AI is neither a panacea nor a threat in itself. Its impact within the DAS classrooms depends on how intentionally it is integrated into existing pedagogical frameworks. Ng, E. N. (2024) reinforces the centrality of educator’s expertise, while Pande, K. and Teng, K. E. (2024) highlights the importance of self-regulation and metacognition in student’s use of AI. Emerging work on RAG in education suggests that retrieval-supported generation can complement these priorities by improving factual grounding and enabling teachers and students to inspect where information originates (Lewis et al., 2020; Huang et al., 2024).​

When used thoughtfully, AI can enhance lesson planning efficiency and support student’s learning as a scaffold. When misused, it risks undermining independence, confidence, and deep understanding. Moving forward, the challenge for DAS educators is not to avoid AI, but to guide its use purposefully in service of inclusive, reflective, and meaningful learning.