None of them on its own. A school leadership team in England starts with the Department for Education's generative AI guidance, because it sets out what the government expects of you. TeachAI's toolkit then gives you wording for a policy, UNESCO's competency frameworks tell you what staff and students should learn, and Jisc's maturity model, written for colleges and universities, gives you a scale of progress.
Sources last checked on November 16, 2026. Every framework below was read at its publisher's own site on that date.
The question is hard only because it is usually asked the wrong way round. These four are not rivals. A government department wrote one, a United Nations agency wrote another, a coalition of nonprofits wrote the third and a sector technology body wrote the fourth, each for a different reader and a different job. A leadership team that treats them as four answers to one question ends up choosing by reputation, and reputation tells you nothing about fit.
The gap leaders need to close is rarely a shortage of frameworks. In a Gallup and Walton Family Foundation survey of 2,069 US teachers, reported by Education Week on May 28, 2026, 82 percent said they had received no formal guidance on how to apply AI tools, and 34 percent said they had received no guidance of any kind. All four of the documents compared here were already public when those teachers answered. Published frameworks do not turn into guidance in a building until someone on the leadership team decides which one does which job.
What does each framework actually give you?
Each gives you one thing: the DfE gives schools in England their boundaries, TeachAI gives policy wording, UNESCO gives a map of skills, and Jisc gives a scale of progress. The table sets out what I found in each when I read them on November 16, 2026.
| Framework | Written by, and for whom | Date shown at source | What you get | What it does not contain |
|---|---|---|---|---|
| Generative artificial intelligence (AI) in education | Department for Education; schools and colleges in England | Updated August 12, 2025 | The government's position on safe use, pupil supervision, age restrictions, personal data and formal assessment | A template policy, stages of maturity, or a list of competencies |
| AI Guidance for Schools Toolkit | TeachAI, led by Code.org, ETS, ISTE, Khan Academy and the World Economic Forum; education authorities, school leaders and teachers worldwide | First released 2023, updated 2025 | Seven principles, a three-part plan of work and an editable sample guidance document | Legal force in any country, or a maturity model of its own |
| AI competency frameworks for teachers and students | UNESCO; countries, curriculum developers and teacher training programs | Published 2024; pages last updated January 16, 2026 | 15 teacher competencies and 12 student competencies, each set out at three levels | A school policy, rules for a school, or a self-assessment for one |
| AI maturity model and toolkit | Jisc; universities, colleges and skills providers in the UK | Model first released 2021, updated March 8, 2024 | Five stages of maturity, with resources for each | Any reference to schools on the pages I read |
Three details in that table deserve a closer look.
The DfE document is a position, not a plan. It applies to England only. Its central sentence hands the choice back to you: "Schools and colleges are free to make their own choices about the most suitable use cases," provided they meet their wider statutory obligations. It tells you pupils should use generative AI only with safeguards such as close supervision, that staff should respect the age restrictions tools set, and it recommends that personal data is kept out of generative AI tools. For formal assessment it points you to the Joint Council for Qualifications. Two companion publications carry the practical detail. The product safety standards, updated January 19, 2026, are "mainly intended for edtech developers and suppliers to schools and colleges" and run from filtering and monitoring through to mental health and manipulation. The support materials, developed by the Chiltern Learning Trust and the Chartered College of Teaching and updated on May 19, 2026 for the 2026 to 2027 academic year, add four training modules for all staff and a set of materials for leaders.
TeachAI is the only one of the four that hands you draft wording. Its seven principles are named purpose, compliance, knowledge, balance, integrity, agency and evaluation. Its plan of work has three areas: guidance and policy, organizational learning, and improvement and transformation. Its sample document, "Guidance on the Use of AI in Our Schools," can be opened and edited. TeachAI is candid about the limits. The sample is "provided for illustrative purposes and is not a definitive model," and the toolkit's examples "are not meant to be copied verbatim."
UNESCO writes for countries. The teacher framework describes itself in these words: "As a global reference, this tool guides the development of national AI competency frameworks." It sets out 15 competencies across five dimensions (a human-centred mindset, the ethics of AI, AI foundations and applications, AI pedagogy, and AI for professional learning) at three levels: acquire, deepen and create. The student framework sets out 12 competencies across four dimensions at three levels: understand, apply and create. Neither tells a school what to permit on Monday morning, and neither was written to.
Which one should you start with? A table by school type
Start with the document your own regulator wrote, then add one framework for each job that document leaves undone. The table below is my suggested way of sorting it, not an official mapping, and you can copy it into a leadership paper as it stands.
| If you lead | Start with | For policy wording | For staff and student skills | For gauging progress |
|---|---|---|---|---|
| A state-funded school or academy in England | The DfE guidance and its materials for leaders | TeachAI's sample guidance, edited to fit | UNESCO's two competency frameworks | Jisc's five stage names, as a prompt for discussion only |
| A multi-academy trust | The DfE guidance, adopted once at trust level, with the product safety standards for procurement | TeachAI's seven principles as the trust-wide spine | UNESCO, with one level chosen for the year | Jisc's stage names, compared school by school |
| A school outside England, including British-curriculum schools abroad | Your own regulator's rules. The DfE guidance is a useful reference with no force outside England | TeachAI, which is written for education systems worldwide | UNESCO, which is written as a global reference | Jisc's stage names, as a prompt only |
| A US school or district | Your state's guidance and your district's existing policies. TeachAI's sample guidance page links to guidance from 26 states, as of January 2025 | TeachAI's sample guidance and principles | UNESCO's two frameworks | The K-12 Gen AI Maturity Tool from CoSN and the Council of the Great City Schools, which TeachAI's toolkit points to and this comparison does not cover |
| A further education or sixth-form college in England | The DfE guidance, which covers colleges | Jisc's strategic framework for colleges, with its three pillars | UNESCO's two frameworks | Jisc's maturity model and toolkit, which were written for you |
A caution on the last column. Jisc built its model to help you "understand where your college or university is on its AI journey," and Jisc's National Centre for AI calls it "a tool to frame conversations." Its five stages are approaching and understanding, experimenting and exploring, operational, embedded, and optimized or transformed. A school can borrow those five names to describe where it stands. It should not score itself against resources written for institutions with very different staffing and systems. College leaders will find more on that in what a college principal decides about AI.
Which framework answers the question you are actually asking?
Match the document to the question on the table in front of you. I call this the Job Test, and I offer it as a working habit, not as research: before you adopt any AI framework, name the job you need done (a rule to stay inside, words for a policy, skills to teach, or a measure of progress), then use the document that was written to do that job.
| The question in the meeting | The job | Use |
|---|---|---|
| What are we allowed to do, and what is expected of us? | A rule to stay inside | The DfE guidance in England; state and district guidance in the US |
| Is this product safe to bring into school? | A rule to stay inside | The DfE product safety standards, turned into questions for the supplier |
| What should our AI policy say? | Words for a policy | TeachAI's sample guidance and seven principles |
| What should we do first, second and third? | Words for a policy | TeachAI's three areas of work |
| What should our teachers be able to do? | Skills to teach | UNESCO's teacher framework, with the DfE modules as training content |
| What should students learn about AI? | Skills to teach | UNESCO's student framework |
| How far along are we? | A measure of progress | Jisc's five stages, in a college; a readiness check built for schools, in a school |
If the question in your meeting is the first one and nobody above you has answered it, I have written separately about what a principal should do when nobody has given them AI guidance. If it is the last one, a seven-point readiness check written for schools will serve you better than a scale borrowed from universities.
The Mistake I See Most Often
The mistake I see most often is a leadership team adopting one framework whole, as though the choice settled the matter. Across the leadership teams I work with, it takes three forms.
The first is the competency list pasted into a policy. UNESCO's fifteen teacher competencies are a sound basis for a training plan. A teacher deciding whether to put a class through a chatbot activity cannot act on them, because they describe what a teacher should become, not what this school permits.
The second is the borrowed policy that was never checked against the regulator. A school in England can adopt every one of TeachAI's principles and still have nothing written down about pupil supervision or personal data, which are the points the DfE guidance actually raises.
The third is the unflattering score. A school measures itself against a scale built for universities and concludes that it is behind, when the scale was never describing a school.
I have advised the UK Department for Education, the Knowledge and Human Development Authority in Dubai and the Ministry of Education in Kazakhstan. Guidance written for a whole system reads the way it does for a reason. A national body writes for thousands of institutions at once. It states a position and leaves the local decisions to you on purpose. That is what "free to make their own choices" means. It is an assignment, and many leadership teams read it as permission to wait.
What do all four frameworks leave out?
All four leave out the decisions that are yours alone: which tools, which year groups, who signs off, and what happens when something goes wrong. None of them will tell you who approves a new tool, so you still need a requirement list for approving AI tools. None of them is your policy, and the gap between a sample and a policy a school can enforce is the subject of what to copy and what to decide in a school AI policy template and, for American readers, the US district version. None of them sequences a year of professional learning, which is the job of a CPD lead's AI training plan.
There is a fair objection here: four documents is a lot of reading for a small school. It is. A one-form-entry primary with no IT team does not need all four. Read the DfE guidance, run the DfE module on safe use with every member of staff, write one page of rules, and leave the rest until a real question sends you to it. The Job Test works in that direction too. With no job to do, you do not need the framework.
These documents also move. The DfE refreshed its support materials in May 2026, TeachAI revised its toolkit in 2025, and UNESCO's pages were last updated in January 2026. Hanover Research, in a report covered by EdTech Magazine on November 5, 2026, recommends that districts keep a "living" AI framework for the same reason. Whatever you build on these four, put a review date on it.
What should a leadership team do this term?
Do five things, in this order.
- Write the question down. Before anyone names a framework, agree on the one question the team needs answered this term, in the words of the third table.
- Read your regulator's document in full. In England that is the DfE guidance. List every expectation in it and mark the ones you cannot yet show you meet.
- Borrow wording, then make it yours. Open TeachAI's sample guidance, keep the sections that fit, and rewrite each one around a decision your school has actually made. A strategy people use is built the same way.
- Choose one level of one competency framework for this year's training. For most staff teams that is UNESCO's "acquire" level. Leave "deepen" and "create" for the staff who are ready.
- Set a review date and check the sources again. Note the date shown on each document today. If one has changed when you return, reread it before you revise anything. The first 90 days of leading AI change gives you a sequence for the weeks in between.
If your leadership team is working out which of these to build on, and what to decide for itself, that is the work I do with schools on AI strategy. I also take this question to trust boards and leadership conferences as an education keynote.
Sources and further reading
- Department for Education, Generative artificial intelligence (AI) in education, policy paper, applies to England, updated August 12, 2025.
- Department for Education, Generative AI: product safety standards, updated January 19, 2026.
- Department for Education, Using AI in education settings: support materials, published June 10, 2025, last updated May 19, 2026.
- TeachAI, AI Guidance for Schools Toolkit, first released 2023, updated 2025, including the principles, the framework and the sample guidance.
- UNESCO, AI competency framework for teachers and AI competency framework for students, 2024, pages last updated January 16, 2026.
- Jisc, AI maturity toolkit for tertiary education and Leading AI in colleges: a strategic framework, no date shown on either page.
- Jisc National Centre for AI, Our AI in Education Maturity Model: an update for 2024, March 8, 2024.
- Lauraine Langreo, Teachers Say Lack of AI Guidance Is a Major Problem, Education Week, May 28, 2026, reporting a Gallup and Walton Family Foundation survey of 2,069 US teachers.
- Kellie Escoto, National AI Research Report Covers How Tools Are Used in Classrooms, EdTech Magazine, November 5, 2026, reporting Hanover Research's K-12 AI Integration Trends.
Dan Fitzpatrick is the founder of The AI Educator, a Forbes contributor, an international keynote speaker and a bestselling author who has trained more than 150,000 educators across more than 30 countries.


