AI literacy for AI natives? A four-country comparison

Synthesis: Using neo-ecological theory and a five-component AI Literacy progression model, this document analysis asks whether early childhood education and care (ECEC) curriculum-guiding documents in Australia, China, Japan and South Korea reflect emerging AI literacy. It finds that early childhood guidance lags far behind AI's spread into children's homes: the Japanese and South Korean curricula and China's national guidelines contain no explicit AI reference, and although Shanghai's strategic action plan folds kindergartens into a citywide AI education agenda, it specifies no learning progressions for young children. Australia shows only limited, uneven alignment, strongest in "Interacting with AI Agents" and entirely absent in "Understanding Human-AI Relationships." Because AI literacy development is left largely to the home, the authors warn that widening differences in children's foundational capabilities may entrench educational disadvantage along AI literacy lines in what some call the "AI arms race." They argue that curriculum documents should articulate explicit, culturally responsive AI literacy priorities and developmental expectations for young children.

Key Findings

1. Japan's and South Korea's national ECEC curricula contain no explicit reference to AI literacy, prioritizing human-to-human relationships, group harmony and offline safety over digital or AI-mediated learning experiences.
2. China's national Early Learning and Development Guidelines contain no AI references; only Shanghai's 2024–2026 action plan positions kindergartens within a citywide AI education agenda, without pedagogic progressions.
3. Australia is the only jurisdiction with any alignment to the five-component AI literacy model, and that alignment is limited, uneven and found only in the non-mandated Early Childhood Australia statement.
4. Australia's strongest alignment is "Interacting with AI Agents"; alignment is weakest on "Understanding Human-AI Relationships," which is absent from every document reviewed.
5. All curriculum-guiding documents analyzed were developed before OpenAI's 2022 release of ChatGPT, so their silence on AI partly reflects timing rather than deliberate policy omission.
6. With no curriculum mandate, young children's AI literacy development remains contingent on the home learning environment, amplifying existing inequities as AI becomes mainstream in society.
7. The authors call for explicit AI literacy learning priorities, developmental expectations and culturally responsive resources to equip educators to facilitate equitable learning.

How the four countries position AI in early childhood curricula

The study asked: To what extent do such documents in Australia, China, Japan and South Korea reflect progressions in AI literacy? Using neo-ecological theory and Yang et al.'s (2026) five-component AI literacy model — understanding foundational AI concepts, creating AI artifacts, interacting with AI agents, developing awareness of AI ethics, and understanding human-AI relationships — the multilingual team analyzed national curricula and supplementary AI policy documents in their original languages or official translations. The headline result is sobering: no single country addressed all five components, and across China, Japan and South Korea alignment was entirely absent. Australia was the only jurisdiction where any alignment appeared, confined to the non-mandated Early Childhood Australia statement, and even there the alignment looks largely incidental, arising from children's everyday digital experiences rather than an explicit positioning of children as developing AI literacy competencies.

Australia: limited and uneven alignment

The Early Years Learning Framework V2.0 (EYLF, 2022) references digital technologies across multiple sections but contains no explicit reference to AI. The supplement is Early Childhood Australia's (2025) Statement on Young Children and Digital Technologies, which does address AI and spans robots, voice assistants, augmented and virtual reality, and coding toys. Its alignment with the five-component model is thin. Component one draws only on the brief phrase "explore how AI works," with no elaboration. Component two rests on children "co-creating prompts for AI with educators," though the interface (voice or text) is unspecified. Component three is the strongest, positioning children as interacting with conversational agents in smart speakers, tablets and phones. Component four presents AI ethics as obligations of educators and services rather than competencies children develop. Component five — human-AI relationships — shows no alignment at all. Across the five components, Australia's average score is minimal alignment, and the guidance is non-regulatory.

Why educational values, not technology, explain the differences

The authors argue that the absence of AI in Japanese and South Korean documents signals educational priorities rather than disengagement with technological change. Japan's human-centered vision of Society 5.0, with its "Convergence Knowledge" and "panoramic perspective," situates technology as serving wellbeing and social sustainability; Japanese ECEC curricula remain anchored in non-digital, hands-on, social-emotional learning and group participation, consistent with mimamoru. South Korea's Nuri Curriculum blends an academic focus with play and emphasizes safety, responsibility, fairness and diversity; an early Robot-Learning pilot using the iRobiQ robot in 2008–2009 was not sustained once the technology matured poorly. China instead treats AI as a strategic priority for national development and innovation, yet even Shanghai's action plan emphasizes institutional capacity — innovation laboratories, smart learning spaces, "experimental schools" — rather than developmental pathways for children's AI literacy.

Why AI literacy is left to the home

Because no country provides a structured roadmap, young children's AI literacy development is likely to remain contingent on the home learning environment. The paper marshals evidence of how uneven that environment is: in Shanghai, smartphone and tablet ownership among families with children aged one to three reaches 96.5% and 89.4% respectively; a United States survey of 1,447 parents found that 44% of children aged three to five own smartphones and 32% own a smart speaker; in South Korea children begin using digital media at an average age of 12–14 months, and while 42.6% of households report rules about media use, roughly 74.0% of those confirm the rules are followed. A recent Ipsos survey of 25,781 European Union citizens found 76% agreed digital skills teaching should start at an early age and 85% that digital skills are necessary to use generative AI skills. Where home microsystems diverge from ECEC microsystems, children face a disjunct rather than continuity, and differences in foundational capabilities may entrench educational disadvantage.

What this means for practice

• Instructors. Treat children's existing AI-related curriculum experiences as funds of knowledge to build on, rather than assuming a uniform starting point or leaving AI literacy to the home.
• Curriculum designers. Articulate explicit, age-banded AI literacy learning priorities and developmental expectations for the preschool years, so that the guidance is not left implicit or non-mandated.
• Policymakers. Close the gap between strategic AI ambitions and early-childhood Pedagogies and Teaching Strategies by specifying what young children should learn and how educators should teach it, not just by funding infrastructure.
• Teacher educators. Embed AI literacy in both pre-service and in-service preparation; without curriculum or workforce guidance, teacher AI competency will remain self-resourced and variable.
• Researchers. Develop and validate AI learning progressions for the preschool years, and study the nature and causes of variability in young children's familiarity with AI.

Limitations

• The analysis examined only publicly available documents; as AI and children's access evolve, responsive policies and curriculum documents are expected to adapt.
• No validated AI learning progressions yet focus on the preschool years, and using Yang et al.'s (2026) deductive framework can only identify what the framework anticipates.
• All analyses used a single large language model within one institutionally governed AI environment, so findings may not transfer to other models or less-governed settings.
• Employing AI in the translation and back-translation process leaves uncertain how fully culturally embedded policy language was captured within an English-language framework.
• The study did not examine how well teachers are prepared to equip children with AI literacies, leaving a significant field for future research.

Citation

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.