Early childhood curricula lag behind the AI shaping children’s lives

Artificial intelligence has moved into children’s homes faster than the documents that guide their earliest education. A new document analysis comparing early childhood education and care curricula in Australia, China, Japan and South Korea finds that none of the four countries embeds a structured approach to AI literacy for children aged three to five, even as voice assistants, smart toys and AI-enabled apps become routine fixtures of family life. The study, published in the journal AI, Brain and Child, examined the national curriculum-guiding documents currently in use in each country and asked a deceptively simple question: to what extent do these documents reflect progressions in AI literacy?

The answer, in three of the four cases, was barely at all. Japan’s three national ECEC frameworks — the National Curriculum Standards for Kindergartens, the Curriculum Standards for Integrated ECEC Centres, and the Guidelines for Care and Education in Nursery Centres — contain no explicit mention of AI. South Korea’s 2019 Revised Nuri Curriculum Commentary likewise showed no alignment with the AI literacy framework the researchers applied. China’s national Early Learning and Development Guidelines for Children Aged 3 to 6 Years contain no explicit references to digital technologies at all. Australia was the only jurisdiction where any alignment emerged, and even there it came from a non-mandated professional statement rather than the national framework itself.

To make the comparison systematic, the team used a five-component AI literacy progression model developed by Yang and colleagues, which to the authors’ knowledge is the first to map AI literacy learning from kindergarten through Year 12. The components cover understanding foundational AI concepts, creating AI artefacts, interacting with AI agents and AI products in various contexts, developing awareness of AI ethics, and understanding human-AI relationships. The analysis was framed by neo-ecological theory, which treats the digital devices children use as objects within their physical microsystems and, in its most recent formulation, positions AI within the child’s virtual microsystem as something the child can interact with reciprocally.

The findings varied sharply by country, and the researchers attribute much of that variation to underlying educational values. Japan’s curriculum documents emphasise human relationships, holistic development and group harmony, consistent with the national Society 5.0 vision that places technological innovation in service of human wellbeing; their silence on AI reflects those priorities rather than disengagement, the authors argue. South Korea’s Nuri Curriculum similarly prioritises physical and naturalistic interactions, even though the country piloted robot-assisted learning in kindergartens as early as 2008 with the iRobiQ robot — an experiment that faltered when the immature technology produced slow, error-prone and contextually inappropriate responses that diminished children’s engagement and teachers’ willingness to persist.

China presents a striking internal contrast. The national ELDG addresses none of the five AI literacy components, and its science and arts learning goals do not transition into digital or computational competencies. Yet Shanghai’s Action Plan for Promoting the Implementation of Artificial Intelligence to Empower High-Quality Development of Basic Education, covering 2024 to 2026, explicitly positions kindergartens as central hubs for innovative AI education. Kindergartens may be designated as AI Education Experimental Schools, develop their own AI curricula, and establish innovation laboratories, smart learning spaces and experiential centres, sometimes in partnership with leading AI enterprises. But the plan operates at the level of institutional capacity rather than pedagogy: it does not articulate developmental expectations for children’s AI learning or educator practices aligned with any learning progression.

Australia’s picture is fragmented. The Early Years Learning Framework V2.0 references digital technologies across multiple sections but contains no explicit mention of AI. Alignment appeared only in Early Childhood Australia’s 2025 Statement on Young Children and Digital Technologies, an authoritative but non-mandated professional document. Its strongest alignment was with the third component, interacting with AI agents, positioning children as users of conversational agents in smart speakers, tablets and smartphones and of AI embedded in toys, games and apps. There was a single brief nod to foundational concepts — an invitation for educators to explore how AI works — one point of alignment on children co-creating prompts with educators, and no alignment whatsoever with understanding human-AI relationships, including algorithmic bias and the fact that AI is programmed and controllable.

Methodologically, the study is notable for its AI-augmented approach. The multilingual research team, all qualified early childhood teachers, analysed documents in their original languages or official translations, using Claude Sonnet 4.5 within a university-governed environment. The researchers required the model to supply source quotations, page numbers and translations, which were then verified by team members with the relevant language proficiency — a human-throughout-the-loop process for which the researchers retain full responsibility. They acknowledge limitations: all analyses used a single model in one institutional environment, and the deductive framework can only identify what the five components anticipate, potentially missing culturally situated conceptions of AI literacy that fall outside them.

Timing matters to the interpretation. Every national curriculum document analysed was drafted before OpenAI’s public release of ChatGPT in 2022, so the absence of AI references is partly an artefact of when the documents were written. The most recent, Australia’s EYLF, was released almost simultaneously with ChatGPT, but its consultative process and drafting predated it. The researchers also searched for subsequent policy statements addressing AI in early childhood, which is how Shanghai’s plan and comparable regional documents entered the analysis, while national-level documents remained the inclusion criterion.

The equity implications are central to the authors’ argument. Young children’s knowledge of AI is largely shaped by digital parenting practices at home, producing wide variability in familiarity before children ever reach a classroom. Data cited in the study illustrate the scale: among Shanghai families with children aged one to three, smartphone ownership reaches 96.5 percent and tablet ownership 89.4 percent, with roughly half of that cohort using devices daily. In the United States, a survey of 1,447 parents found that 44 percent of children aged three to five had smartphones and 32 percent owned smart speakers. In South Korea, children begin using digital media at an average age of 12 to 14 months. Without curriculum guidance, the authors contend, children’s AI literacy will remain contingent on the home environment, where they may interact with AI agents and create artefacts far more often than they develop awareness of ethical considerations and risks — entrenching educational disadvantage along the lines of AI literacy as the technology becomes mainstream.

The researchers conclude that none of the four nations has a structured roadmap for supporting early AI literacy, and they call for curriculum-guiding documents to articulate explicit AI literacy learning priorities and developmental expectations for young children, accompanied by culturally responsive resources for educators. Shanghai’s inclusion of foundational AI concepts in strategic planning that encompasses kindergartens is flagged as a significant precedent for future policy reviews in China. The team also notes open questions for future research: whether teachers are prepared to equip children with AI literacies, whether countries that integrate AI literacy early better prepare children for a digital society, and how variable AI fluency within homes and classrooms might feed back into national education values. For now, the documents that guide the earliest years of schooling remain anchored in a pre-generative-AI world.

Subject of Research: A document analysis of early childhood curricula in Australia, China, Japan and South Korea examining the extent to which they reflect AI literacy progressions for young children.

Article Title: AI literacy for AI natives? A four-country comparison

Article References: Cohrssen, C., Madanipour, P., Guarrella, C., Chen, M. Y., & Park, S. (2026). AI literacy for AI natives? A four-country comparison. AI, Brain and Child, 2(1), Article 21. https://doi.org/10.1007/s44436-026-00045-3

Image Credits: AI Generated

DOI: 10.1007/s44436-026-00045-3

Keywords: AI literacy, early childhood education, curriculum policy, Australia, China, Japan, South Korea, document analysis, educational equity, generative AI, learning progressions, ECEC