In a world where individuals need to continuously adapt to evolving digital ecosystems, learning technologies play a key role in helping learners rapidly understand, evaluate, and apply complex technical concepts. To meet this need, advances at the intersection of computer science and learning sciences are driving the creation of intelligent systems that enhance digital literacy. From a computer science perspective, Large Language Models (LLMs) offer new opportunities for tutoring, feedback, and conceptual support, yet their tendency to generate unverified information risks undermining understanding and epistemic trust. From a learning perspective, effective educational technologies are required to ensure accuracy, transparency, and cognitive alignment with learners’ needs. Retrieval-Augmented Generation (RAG) is emerging as an effective method to mitigate hallucinations by grounding responses in external sources, but current implementations often rely on simplistic retrieval methods that neglect the hierarchical and pedagogical structure of educational materials. This doctoral research investigates how RAG-based methods can be tailored and integrated into educational activities to foster digital literacy. The work focuses on developing a generative assistant capable of semantically retrieving domain materials, synthesizing authoritative information, and producing pedagogically aligned explanations. The expected contributions are theoretical, by advancing hybrid retrieval and explainability mechanisms tailored to learning contexts, and practical, by demonstrating a literacy-oriented assistant that bridges technical content and learner comprehension. As a case study, the approach will be applied to cloud literacy, contextualizing service operations and costs via authoritative documentation and real-world data.