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.
The short version
- What happened
- UNESCO IITE systematically analysed 36 publicly available national, institutional and academic frameworks for teacher AI competence.
- Who it affects
- Teachers need protected time, appropriate assessment and practical support, not another unfunded list of expectations.
- What remains unproven
- The denominator is 36 framework documents, not a sample of teachers, schools or pupils.
Keep in mind
The denominator is 36 framework documents, not a sample of teachers, schools or pupils.
The review compares content and implementation design but does not measure classroom outcomes or causal effects.
Publicly available frameworks may not represent unpublished or locally implemented approaches.
Source analysis · 2 linked sources. Primary analysed source date: 28 September 2026. Earliest linked supporting source: 8 August 2024.
Read the full report · 4 min The evidence, methods, implications and limitations
A map of 36 frameworks—not another single syllabus
The UNESCO Institute for Information Technologies in Education published Teacher AI Competency Frameworks on 28 September with Shanghai Open University. The report systematically analyses 36 publicly available frameworks issued by national education authorities, international policy organisations, institutions and academic researchers. Its aim is to help policymakers design national approaches rather than ask every school to adopt one universal list unchanged.
The analysis follows four stages: identifying the aspects of teacher AI competence covered in research and practice; examining implementation at strategic, tactical and operational levels; comparing approaches to evaluation and assessment; and developing design recommendations. The annex records the frameworks examined, while the conclusion provides a checklist. The unit of analysis is therefore a framework document, not a teacher, school, lesson or pupil.[1][2]
Competence has to include judgement, not only tool use
The report treats AI literacy as more than prompt writing. Teachers need enough technical understanding to recognise what a system can and cannot infer, ethical understanding to protect rights and privacy, pedagogical judgement to decide whether use advances a learning goal, and professional capacity to evaluate outputs and keep learning. That direction is consistent with UNESCO's 2024 framework, which organised 15 competencies across a human-centred mindset, ethics, foundations and applications, pedagogy and professional learning.
For a school, the practical implication is that access to a chatbot is not a training programme. Staff need examples tied to subjects and age groups, time to examine failure cases, clear rules for pupil data and assessment, and a route to decline a tool that does not fit the lesson. Leaders also need to distinguish competence to teach about AI from permission to use a particular product. A confident user can still be operating under a weak procurement or safeguarding policy.[1][2]
Implementation and assessment are the hard part
UNESCO's three implementation levels make responsibility visible. At the strategic level, governments set goals, rights and funding. At the tactical level, curriculum bodies and teacher-education providers translate those goals into programmes, standards and support. At the operational level, teachers apply them in classrooms. If a policy names competencies but supplies no curriculum time, assessment method or professional support, the burden falls on individual teachers and provision becomes uneven.
Assessment should also measure performance rather than familiarity with product names. A useful task might ask a teacher to identify fabricated evidence, adapt an AI-supported activity for a learner with additional needs, explain a data-protection choice and decide when not to use AI. Self-confidence surveys can help diagnose demand, but they do not show that someone can handle a consequential classroom situation. National frameworks should publish progression criteria and make room for local language, infrastructure and curriculum conditions.[1]
What the review cannot tell schools
A review of 36 documents can reveal coverage and gaps, but it cannot identify the best framework by measured impact. The report does not compare randomised training programmes, observed teaching practice, pupil attainment, workload, inclusion or safeguarding incidents. Publicly available frameworks may also over-represent systems with the resources and language reach to publish formal documents. A well-designed framework can remain inactive if teachers lack devices, connectivity, time or trusted professional learning.
Our assessment would strengthen if education systems published implementation studies showing which training and assessment approaches change classroom decisions over time, with results separated by subject, school phase, language and resource level. It would weaken if frameworks became compliance checklists detached from teaching quality. For now, the review gives policymakers a serious map of what to include and where responsibility sits; it should begin local testing, not end the debate about what teachers and learners need.[1]