If you are a school leader trying to work out your school's approach to AI literacy, the first thing you will discover is that there are too many frameworks and not enough plain English explanations of what they actually say.
This post fixes that. No jargon. No acronyms without explanation. Just what each AI literacy framework covers, where they agree, and what it means for your curriculum, your governors, and your next Ofsted conversation.
Five frameworks in 1 minute
There are five AI literacy frameworks that matter for UK and international schools right now. They come from different organisations but converge on the same core idea: AI literacy is a cross-curricular competency, not a Computing topic.
The five are UNESCO, the OECD/EC AILit Framework, PISA 2029, the US Department of Labor, and Anthropic's four-dimensional model. Regional frameworks exist too (the UAE's mandatory K-12 curriculum, DigComp 3.0 for EU schools) but these five form the global baseline.
AILitKit maps every activity it generates to all five of these automatically. You do not need to read the framework documents yourself. But understanding what they say helps you explain to governors, parents, and inspectors why AI literacy matters and what your school is doing about it.
UNESCO: AI Competency Framework for Students
Published 2024. The broadest and most internationally recognised.
UNESCO organises student AI competency into three areas. A human-centred mindset (understanding AI as a human-made tool with human biases embedded in it). AI techniques and applications (what AI can do, how it works conceptually, where it is used). And the ethics of AI (bias, fairness, privacy, social impact).
Within each area, students progress from understanding, to applying, to evaluating and creating. A primary student might recognise that a recommendation algorithm exists. A secondary student might evaluate whether it treats all users fairly.
UNESCO matters for international schools especially. If your school serves families from multiple countries, UNESCO is the framework parents are most likely to have heard of. It also maps well to the other frameworks, so teaching to UNESCO coverage means you are building competence that counts elsewhere too.
OECD/EC AILit Framework: the one that shapes assessment
Published as a review draft in May 2025. Final version expected 2026. This is the framework to watch because it directly informs the PISA 2029 assessment.
AILit organises AI literacy into four domains. Engage with AI covers recognising AI, understanding how it works, and evaluating its outputs. Create with AI covers using AI tools to generate, refine, and iterate. Manage AI covers responsible use, privacy, bias, and knowing when not to use AI. Design AI covers understanding the human choices behind AI systems, who benefits, who is excluded, and what responsible development looks like.
Twenty-two competences sit across these four domains, each with learning scenarios for primary and secondary classrooms.
For curriculum planning, the four-domain structure gives you a map. If your school's AI literacy activities all fall under "Engage with AI" and nothing under "Design AI," that is a gap you can see and fill. AILitKit's Topic guides map an entire unit against these four domains, showing which lessons develop which competences. Whole Curriculum guides do the same across subjects, and the coverage heatmap on higher tiers makes the distribution visible at a glance.
PISA 2029: why this stops being optional
Every three years, the OECD assesses 15 year olds across member countries. The results shape national policy, influence funding, and drive curriculum reform. In 2029, PISA adds a new domain: Media and AI Literacy, or MAIL.
Students will be tested on understanding algorithms, evaluating AI-generated content for credibility, identifying bias, and making ethical decisions about AI use.
For UK schools, this means AI literacy moves from "a good idea" to "something our 15 year olds are assessed on internationally." Schools that start building these competences now have four years. Schools that wait will have weeks.
For international schools competing on academic outcomes, PISA performance is a differentiator. Being able to tell prospective families "our students are prepared for the PISA 2029 AI literacy assessment" carries weight.
This is one of the reasons AILitKit generates a governor report. A governing body needs to understand where the school stands on AI literacy readiness, not in abstract terms, but in terms of framework coverage, identified gaps, and next steps. The governor report translates classroom activity into governance language.
US Department of Labor: the most practical framework
Published February 2026. Voluntary guidance, but it signals the direction of federal funding for workforce and education programmes.
The DOL framework has five content areas. Understand AI Principles (what AI is, how it works, its limitations). Explore AI Uses (real-world applications across industries). Direct AI Effectively (prompting, context-setting, iterating). Evaluate AI Outputs (accuracy, completeness, bias, fitness for purpose). Use AI Responsibly (data privacy, workplace policies, accountability).
What makes the DOL framework distinctive is its focus on doing. It is the only national framework that names prompting, "Direct AI Effectively," as a foundational competency. Most frameworks are about understanding AI. The DOL adds the practical skill of communicating with it clearly.
For UK schools, the DOL is useful as a reference point when parents or governors ask "but what does AI literacy actually look like day to day?" The five content areas are plain, practical, and easy to explain. They also map cleanly to the other frameworks: DOL's "Evaluate AI Outputs" is the same competency as PISA's "Evaluating Credibility," Anthropic's "Discernment," and OECD AILit's "Engage with AI."
Anthropic 4D: the classroom-ready model
Anthropic, the company behind the AI model Claude, describes AI literacy through four dimensions: Description (giving AI clear instructions), Discernment (evaluating what AI produces), Delegation (knowing when to use AI and when not to), and Diligence (using AI responsibly and safely).
A note on sourcing: unlike the other frameworks in this list, the Anthropic 4D model is not a standalone published policy document. It comes from Anthropic's education research and thinking about what makes a competent AI user. AILitKit uses it because of its practical clarity. It answers a focused question: what does someone need to be good at to use AI well?
For teachers, the 4Ds map directly onto activities. A prompting exercise builds Description. A fact-checking exercise builds Discernment. A debate about when AI should and should not be used builds Delegation. A discussion about data privacy builds Diligence.
Every activity in an AILitKit guide is tagged against these four dimensions alongside the other frameworks. When a school generates multiple guides across subjects and key stages, the tracking on higher tiers shows which dimensions the school is strong on and which need attention.
Where all five agree
These frameworks are not competing. They are converging. Three things appear in every single one.
Evaluating AI outputs. The ability to look at something AI has produced and ask "is this accurate, complete, and fair?" appears in all five frameworks. It is the most universally agreed-upon AI literacy skill. UNESCO calls it Ethics of AI at the evaluate level. OECD AILit calls it Engage with AI. PISA 2029 calls it Evaluating Credibility. DOL calls it Evaluate AI Outputs. Anthropic calls it Discernment. Five names. One competency.
Responsible use. Data privacy. Knowing what not to share. Following policies. Maintaining accountability. Every framework includes this. UNESCO, OECD AILit, DOL, and Anthropic all have a dedicated area for it.
Conceptual understanding. Not coding. Understanding that AI learns from patterns in data, can produce confident but wrong outputs, reflects its training data, and was designed by humans who made choices. This foundational layer appears everywhere and requires no technical knowledge to teach.
And all five agree on one structural principle: AI literacy should be embedded across subjects, not taught as a standalone course. The OECD says educators should "embed AI literacy when and where it aligns with their subject and context." The DOL says learning should be "embedded in context." UNESCO's scenarios span subjects. This is the principle AILitKit is built on.
The vocabulary problem underneath the frameworks
Every framework uses terms like algorithm, training data, bias, hallucination, and model. None of them tell a Year 3 teacher how to explain "algorithm" to a seven year old.
A school that introduces AI vocabulary inconsistently across departments ends up with students who think "bias" means one thing in History and something different in Computing. The frameworks assume a shared vocabulary. In most schools, that vocabulary does not exist.
AILitKit's AI glossary defines every key term at five levels, from KS1 to KS5. Each definition includes a plain explanation and a subject-specific example matched to the key stage. "Training data" in a Year 4 Science guide is explained differently from "training data" in a GCSE Business Studies guide. Both are accurate. Both connect to the same underlying concept. For schools using the Whole Curriculum guide, the glossary becomes a school-wide vocabulary reference that every department can share.
What to actually do
You do not need to pick one framework. You need to start teaching AI literacy, and any framework will tell you you are on the right track.
Start with one subject. One unit. One lesson. Find where AI literacy connects naturally to what you already teach. Build from there. A Lesson guide is the first step. A Topic guide maps the full picture for a unit. A Whole Curriculum guide maps the picture for the school and gives your governors something concrete to discuss.
If you want to understand the frameworks in more depth, these are the source documents:
- UNESCO AI Competency Framework for Students (2024)
- OECD/EC AILit Framework review draft (May 2025) at ailiteracyframework.org
- PISA 2029 MAIL assessment framework at oecd.org
- US DOL AI Literacy Framework (February 2026) at dol.gov
- DigComp 3.0 (November 2025) for EU-specific digital competence
The frameworks are the destination. Your curriculum is the road. The question is not which framework to follow. It is how to start walking.
AILitKit maps every activity to UNESCO, OECD AILit, PISA 2029, DOL, and Anthropic 4D automatically. Higher tiers include coverage heatmaps, governor reports, and tracking across subjects and key stages. Try a free Lesson guide at ailitkit.com