What Counts as AI Literacy? A Comparative Analysis of International Frameworks and System-Level Policies, 2024–2026

Between 2024 and 2026, AI literacy moved from a loosely used phrase to a formal object of international policy. UNESCO, the Digital Education Council, the European Commission and OECD, the OECD's PISA pro‐ gramme and the Council of Europe each published a framework or stand‐ ard defining it. In the same period, school systems began acting on the idea at scale: China set national targets for AI education at every stage of schooling, Boston Public Schools launched AI literacy across its high schools, and New York City paused student-facing generative AI for its younger pupils. This article compares five frameworks and three system- level policies, drawing on a practitioner newsletter series written for school leaders over the same period. It finds strong agreement on four elements: understanding how AI works, critically evaluating its outputs, using it ethically, and keeping human judgement and agency at the centre. It finds five real differences: how much weight goes on the human dimen‐ sion compared with tool use, whether learners should create and shape AI, whether literacy is general or domain-specific, how it develops with age, and whether choosing not to use AI counts as a competence. The policies show that the hardest problems lie outside the definitions, in teacher capacity and in assessment that can tell knowing about verifica‐ tion from actually verifying. We propose a synthesis model and set out implications for curriculum designers, school leaders and assessment bodies.