K-12 AI literacy frameworks after generative AI: A scoping review

Abstract

Artificial intelligence (AI) literacy has become a priority in K-12 education, yet frameworks for teaching and assessing it remain fragmented across grade levels, national contexts, and disciplinary traditions. This scoping review maps and synthesizes framework-level contributions to K-12 AI literacy published between January 2024 and December 2025. Following JBI guidance and PRISMA-ScR reporting standards, we searched four databases and grey literature sources, screened 204 unique records, and applied a purpose-built checklist covering five criteria: grade bands, competencies, critical constructs, assessment, and exemplars. Fifty studies met the inclusion threshold. The corpus spans 15 countries, with the United States (32%), Hong Kong (26%), and mainland China (12%) accounting for 70% of included studies. Frameworks defined AI literacy through cognitive, affective, behavioral, ethical, and social dimensions, but dimensional models used inconsistent labels and constructs, with limited crosswalk among them. The ABCE framework offered the closest approximation to a shared measurement tradition, with adaptation across three educational systems. Assessment and curriculum also developed largely in parallel: scale-development papers reported psychometric evidence without instructional sensitivity data, while curriculum papers assessed learning through lesson-linked tools that limited cross-study comparison. Generative AI competencies, including output evaluation, hallucination detection, and prompt design, remained peripheral in most frameworks despite their relevance to students’ everyday encounters with AI. Geographic concentration and limited attention to socioeconomic status, disability, and linguistic diversity further constrained transferability. The field needs stronger convergence through shared vocabulary, curriculum-aligned assessment, longitudinal evidence, and cross-context validation.