Do early-years curricula teach AI literacy?
Mostly no. A four-country document analysis found no explicit progression across five AI-literacy components in national early-childhood curricula. Australia showed limited alignment only in a non-mandatory statement, while regional Chinese initiatives focused more on infrastructure than children’s competencies.
Editorial responsibility: The Impact of AI Editorial Desk · Report a factual concern
At a glance
- 1The researchers analysed national early-childhood curriculum documents and related current guidance across four Asia-Pacific countries against five proposed AI-literacy components.
- 2No country covered all five components; the paper found no alignment in the national materials examined for China, Japan or South Korea, and only limited alignment in a non-mandatory Australian statement.
- 3This is a multilingual document analysis, not evidence about what teachers actually teach, what children know or whether an AI-literacy curriculum improves outcomes.
Research topic
Whether early-childhood curriculum documents in Australia, China, Japan and South Korea set explicit, developmentally sequenced expectations for children's AI literacy
The answer: policy is behind children’s everyday exposure
A peer-reviewed comparison of early-childhood curriculum documents in Australia, China, Japan and South Korea finds little explicit guidance for building young children’s AI literacy. None of the four settings covered all five elements in the analytical framework: understanding basic AI concepts, creating AI artefacts, interacting with AI agents, recognising ethical issues and understanding relationships between people and AI. The national materials examined for China, Japan and South Korea showed no alignment with that full progression. Australia showed limited alignment only through a professional statement that is not a mandatory national curriculum.
That finding matters because children already encounter recommendation systems, conversational agents, smart toys and generated media before formal schooling. Yet the study does not show that young children should spend more time using AI products, nor that classroom technology produces better learning. Its narrower message is that policy documents rarely tell educators what age-appropriate knowledge, questioning and ethical judgement should develop when AI is already part of children’s environment.[1]
The Impact Brief · Free
Follow the evidence in education.
Get a five-minute weekday briefing on what changed, why it matters and where the evidence comes from. Choose the topics you care about.
What the researchers actually compared
The multilingual team conducted a document analysis rather than a survey or classroom trial. Qualified early-childhood educators on the author team reviewed core curriculum documents in their original languages or official English translations. They also searched in English, Chinese, Japanese and Korean for later policy statements and guidance that addressed digital technology or AI. Their main scope was national-level material currently used for children aged roughly three to five, although they also discussed five influential Chinese municipal or provincial initiatives because those jurisdictions have been leading curriculum development.
Documents were coded against a previously published five-component learning progression. The method gives readers a consistent comparison across policy systems, but it also imposes one framework on culturally different curricula. A document may develop agency, tool awareness or critical thinking without naming AI, and those capabilities could be missed if they do not fit the five predefined components. The authors explicitly acknowledge that an inductive or culturally specific analysis might surface other forms of relevant learning.[1]
Australia had examples, but not a complete progression
Australia’s national Early Years Learning Framework refers to digital technologies but not explicitly to AI. The researchers found partial alignment in Early Childhood Australia’s 2025 Statement on Young Children and Digital Technologies. It mentions exploring how AI works, educators and children co-creating prompts, encounters with conversational agents and AI embedded in toys or apps, and shared assessment of information quality. Those examples touch foundational concepts, creation, interaction and aspects of ethics.
The gaps are as important as the examples. The statement does not specify a developmental sequence or what successful learning looks like, and ethical obligations are directed mainly at educators and services rather than framed as competencies children should develop. The analysis found no alignment with the human–AI relationship component, which includes readiness, agency, algorithmic bias, social benefit and the idea that people design and control systems. Because the statement is advisory, its presence also does not show consistent classroom provision.[1]
China’s policy activity did not translate into preschool expectations
China’s 2012 national early-learning guidelines contain no explicit reference to digital technologies or to the five AI-literacy components used by the study. The authors additionally examined AI education plans from Beijing, Shanghai, Guangdong, Sichuan and Chongqing. Only Shanghai’s 2024–2026 action plan explicitly brought kindergartens into its AI strategy, encouraging teaching practice, experiential activities, innovation laboratories and partnerships with technology companies.
But institutional activity is not the same as a learning progression. The Shanghai plan concentrates on system capacity and facilities rather than specifying what a three-, four- or five-year-old should understand, how an educator should recognise progress, or how children should question generated output. Japan’s early-childhood guidance remained centred on hands-on, non-digital experience, while South Korea’s 2019 Nuri curriculum emphasised physical and naturalistic interaction despite earlier robot pilots. The paper therefore describes a curriculum gap, not an absence of technology or innovation projects.[1]
What teachers and families can use now
The practical implication is not to hand preschool children a chatbot and call that literacy. Educators need concrete, developmentally appropriate ways to help children notice when a response comes from a machine, ask where information came from, compare an answer with trusted human knowledge and understand that an app was made by people with choices and incentives. Those activities can be conversational, playful and low-screen. They should also respect language, disability, local culture and the uneven access children bring from home.
Policy makers can make expectations useful by linking each aim to examples, educator training and observable learning rather than broad calls for innovation. They also need safeguards for privacy, commercial influence and age-inappropriate systems. If guidance simply tells settings to adopt AI, better-resourced centres may buy products while educators remain uncertain about purpose and risk. A curriculum progression should support professional judgement and children’s agency, not turn early childhood into product training.[1]
What the study cannot establish—and what comes next
No children, parents, teachers or classrooms were sampled, so the study cannot estimate how often AI is used, what children understand or whether written curricula match practice. Country comparisons also reflect the documents selected and the framework used. The authors used an LLM during translation and back-translation, then had the multilingual team review and refine the results; culturally embedded policy language may still resist direct translation. The paper reports no dataset because its evidence is public documents rather than participant records.
The assessment would change with observational studies linking written policy to teaching, interviews across urban and rural settings, and trials of age-appropriate activities that measure durable understanding rather than device fluency. Researchers should test whether children can recognise generated content, explain human control, identify uncertainty and transfer those ideas to new systems. Until then, the new paper is strongest as a policy audit: it identifies missing explicit progression, but it does not prove that any proposed curriculum—or any AI product—will benefit a child.[1]
What this means for people
- Early-years educators need examples and training that build children’s judgement without requiring more screen time or product use.
- Families should not interpret everyday device familiarity as evidence that a child understands how AI works, what it may get wrong or who controls it.
- Policy makers can reduce inequality by funding culturally and linguistically responsive educator support, not only infrastructure or vendor partnerships.
Global context
The four settings span different early-childhood philosophies, languages and policy systems across the Asia-Pacific region. The analysis improves regional comparison but does not represent every country or establish a universal curriculum model.
What the evidence does not yet show
- The study analysed documents, not children, educators, teaching practice or learning outcomes.
- A single five-component framework may miss culturally specific or implicitly expressed capabilities.
- The scope was primarily national guidance, with selected sub-national Chinese initiatives added for context.
- LLM-assisted translation and back-translation were reviewed by the multilingual team but may not capture every policy nuance.
- Published guidance does not show whether implementation is consistent, resourced or effective.
What to watch next
- National early-years revisions that define explicit, age-appropriate AI-literacy progressions
- Observational studies comparing written curriculum with classroom practice
- Teacher training and resources that do not depend on commercial AI products
- Child-centred outcome measures covering uncertainty, agency, ethics and transfer
Living evidence record
Impact record IAI-03J83VU
Evidence stage
Studied
Confidence
Supported
Reporting basis
Source analysis
Independent or research support
Present
Record status
Monitoring
Last checked
13 October 2026
Source trail
1 direct source across 1 source type.
People impact
Documented in this record.
Uncertainty
Limits and next checks are explicit.
Stages describe the evidence available—not whether a technology is good or bad. See the public method.
Single-source reporting disclosure
This record analyses one direct source. It can establish what AI, Brain and Child published or reported, but it is not independent corroboration of every performance claim or predicted outcome. The confidence label will change only when broader evidence is added.
Evidence trail
Sources used for this report
Links checked 13 October 2026
This report is labelled source analysis. We summarise and analyse source material in our own words; company statements remain attributed claims until independently supported. Translated summaries preserve the meaning of the original source and link back to it. Read our editorial standards.
Continue the story
Related reporting
Education
Did a global AI course prove clinicians learned more?
Not yet. In a peer-reviewed evaluation, 63 completers from 12 countries reported large gains in confidence after an eight-week course, but there was no objective knowledge test, control group, retention measure or patient outcome.
7 min · 1 source
Education
Can AI make quieter classroom contributions visible? What Japan's Science Council actually proposes
A Japanese Science Council report proposes using generative AI to map classroom discussion and expose bias. Its exploratory survey covered 217 university teachers in Japan and South Korea, but the classroom system and its effects have not yet been established.
6 min · 2 sources
Education
What should teachers learn about AI? UNESCO reviews 36 competency frameworks
A new UNESCO IITE analysis compares national, institutional and academic frameworks for teacher AI competence. It offers a policy checklist, but does not test whether any framework improves classroom outcomes.
5 min · 2 sources
Reader commentary
Add evidence, experience or a question
No account is required. Reader notes are published after a brief civility, relevance and safety check; disagreement is welcome.
Explore commentary across the portal →Published reader notes
0No published reader notes yet. You can start the evidence-led discussion above.
Prefer a private correction or response? Contact the newsroom.