Artificial Intelligence (AI) literacy is becoming an essential competency across disciplines and stakeholders. However, AI-related education poses important difficulties, as it combines many concerns such as technical skills, ethical/legal issues, critical thinking, etc. These difficulties, combined with the variety of learners background, push towards customizable learning support. Such needs could be better addressed through machine processable representations, as those enable distribution, reuse, and automation mechanisms both from Model-Driven Engineering (MDE) and Generative AI techniques. This paper proposes a framework powered by MDE methodologies to support the conceptualization of the AI literacy domain in terms of competencies and learning scenarios. The conceptualization is realized by means of corresponding metamodels in close collaboration with pedagogy domain experts. The expressiveness of the metamodels is demonstrated in a proof-of-concept developed for AI literacy teaching.