No framework in this field answers every question a school has about AI, and the most common implementation mistake is reaching for one at the wrong altitude. A district safety framework cannot tell a history teacher whether students may use AI to outline an essay, and a five-level permission scale cannot supply a coherent AI-literacy curriculum. Sorted by the decision each one governs and how close it sits to a classroom, the named frameworks below fall into five families. The most useful approach is a stack drawn from more than one of them.

Five framework families and a classroom-proximity scale

Frameworks differ mainly in the unit they govern — the system, the district, the curriculum, the teacher, the assignment, or a single act of learning — and none of the five families substitutes for another.

Family	Governs	Where it stops
Governance	Privacy, safety, fairness, transparency, accountability, procurement	Necessary for institutional legitimacy; insufficient for designing one assignment
Readiness	Whether a district has leadership, policy, infrastructure, professional learning to proceed	Whether a rollout can be supported; not what to teach
Competency	What students or teachers should understand and do	Supports curriculum maps and professional learning; may not say what a student may do on tomorrow’s essay
Permission	What AI may do during a task	One assignment-wide label conceals brainstorming, outlining, drafting, editing, citing, verifying
Learning design	Where students should struggle, who performs each cognitive operation, what evidence supports a valid inference	Closest to practice; reaches no further than the lesson

the five framework families

A framework’s classroom proximity says how directly it can guide a Monday-morning decision.

Proximity	Operates mainly at	Typical user
1	National, legal, ethical, procurement, or system governance	Policymaker, superintendent, technology or legal leader
2	District and school readiness, policy, infrastructure, professional learning	District office, principal, instructional-technology leader
3	Curriculum, competencies, standards, and professional development	Curriculum director, department chair, teacher educator
4	Course, unit, assessment, and assignment design	Classroom teacher, instructional coach
5	Immediate classroom routines, prompts, protocols, and student-facing directions	Classroom teacher and students

classroom-proximity scale

Nothing about a proximity of 1 makes a framework inferior to one sitting at the classroom end. A child-rights framework and an assignment checklist solve different problems, and a school that adopts one believing it has answered the other ends up with a policy nobody can apply. That is what a wrong-altitude choice costs, and it costs it quietly: the document is real, the committee met, and the teacher writing an essay prompt in August still has no rule to follow.

The framework matrix

Student AI-literacy and curriculum frameworks

Student AI-literacy frameworks define what students should understand about AI and how they should use, evaluate, and shape it. One domain below bears directly on assignment design: Managing AI, from the OECD–European Commission’s AILit framework, covers deciding deliberately how work should be divided between people and AI systems.

Framework	Core structure	Primary audience	Proximity	Best use	Important limitation
UNESCO AI Competency Framework for Students	Human-centered mindset, ethics, AI techniques and applications, AI system design; Understand, Apply, Create	National systems, curriculum developers, leaders	3	Designing a broad AI-literacy curriculum and progression	Global and general; gives no assignment-permission rules
OECD–European Commission AILit Framework	Engage with AI, Create with AI, Manage AI, Shape AI	Primary and secondary educators, education authorities	3–4	Connecting AI literacy to human–AI task allocation	New framework; implementation evidence limited
Digital Promise AI Literacy Framework	Understand, Evaluate, Use; human judgment and justice central	PK–12 educators, leaders, curriculum designers	3–4	Building practical and critical AI literacy across subjects	Needs local grade-band sequencing and task design
aiEDU AI Readiness Framework	Student competencies, educator competencies, school-leader rubric, district rubric	Districts, schools, teachers, students	2–3	Aligning district readiness to classroom learning	Implementation-oriented, not a detailed curriculum
AI4K12 Five Big Ideas	Perception; representation and reasoning; learning; natural interaction; societal impact	Standards writers, computer-science and AI educators	3	Teaching how AI systems work across K–12 grade bands	Stronger on AI concepts than on generative-AI assignments
ETS AI Literacy Framework	Foundational knowledge, applications, societal implications, ethics, collaboration, critical evaluation, evidence-centered design	Researchers, assessment designers, curriculum developers	3	Developing assessable AI-literacy constructs	A research proposal, not a validated turnkey curriculum
NCTE (National Council of Teachers of English) Working ELA AI Framework	Critical use and examination of AI in English language arts; assignment guidance	Grades 6–12 ELA teachers, departments, curriculum leaders	4–5	Integrating AI literacy with writing and source evaluation	Explicitly provisional; Google.org-supported
AHA Guiding Principles for AI in History Education	Historical thinking, expertise before reliance, explicit policy, disclosure, iterative review	History teachers, departments, curriculum leaders	3–4	Setting discipline-specific principles for history instruction	Principles, not a grade-banded competency sequence
Stanford CRAFT	Free interdisciplinary AI-literacy activities, from brief routines to multi-day sequences	High-school teachers across subjects	5	Immediate classroom implementation and adaptation	A resource collection, not a standards framework
Long and Magerko AI Literacy Framework	Competencies and design considerations for understanding and interacting with AI	Researchers, curriculum and informal-learning designers	2–3	Conceptual foundation under later AI-literacy frameworks	Academic; not written as a secondary implementation guide
Ng and colleagues’ AI-literacy synthesis	Know and understand; use and apply; evaluate and create; ethical issues	Researchers and curriculum developers	2–3	Comparing broad conceptual dimensions of AI literacy	A high-level review, not a classroom system

Student AI-literacy and curriculum frameworks

Teacher-competency and instructional-knowledge frameworks

Two of the frameworks in this family are routinely confused. TPACK — technological, pedagogical, and content knowledge — was revisited for generative AI by Mishra, Warr, and Islam, who argue that the technology puts new pressure on the model because it is unusually adaptable, opaque, unstable, generative, and socially interactive.

SETI, the socio-ecological view of technology integration developed by Helen Crompton, Diane Burke, Christine Nickel, and Agnes Chigona, is not Punya Mishra’s replacement for TPACK. TPACK-XK asks whether the teacher understands enough to design a sound use; SETI asks whether the surrounding school, community, policy, culture, and support systems make that use sustainable.

Framework	Core structure	Primary audience	Proximity	Best use	Important limitation
UNESCO AI Competency Framework for Teachers	Human-centered mindset; ethics; foundations, applications; AI pedagogy; professional learning — Acquire, Deepen, Create levels	Teachers, teacher educators, ministries	3	Designing a coherent teacher professional-learning progression	Broad; teachers need assignment-level design tools
aiEDU Educator Competencies	Teacher knowledge, use, evaluation, facilitation, readiness inside a district-to-student system	Teachers and instructional leaders	3–4	Building teacher capacity within a larger implementation plan	Depends on local professional development and translation
AI for Education SEE Framework	Safe, ethical, effective use, built on knowledge, productive mindsets	Teachers, leaders, instructional designers	4	A memorable professional-learning and decision framework	From a professional-services organization; not a public standard
TPACK revisited for generative AI (TPACK-XK)	Content, pedagogy, technology, and expanded contextual knowledge	Teachers, teacher educators, researchers	3–4	Judging whether a proposed use fits content, pedagogy, context	An analytic lens, not an implementation sequence
SETI (socio-ecological technology integration)	Socio-ecological view across classroom, school, community, policy, culture, infrastructure, norms	Leaders, researchers, implementation teams	1–2	Diagnosing why a sound classroom use succeeds or fails	Too high-level to set permissions or classroom routines

Teacher-competency and instructional-knowledge frameworks

Governance, policy, and district-readiness frameworks

UNICEF’s 2025 guidance cautions against AI that displaces or obstructs children’s independent cognitive and socio-emotional development. Over-assistance, in other words, is named as a hazard by a body that sets no assignment rules at all.

Framework	Core structure	Primary audience	Proximity	Best use	Important limitation
UNICEF Guidance on AI and Children	Child rights, safety, privacy, fairness, transparency, accountability, well-being, inclusion	Governments, regulators, providers, system leaders	1	Establishing a child-rights foundation for AI policy	Not designed for instructional methods
TeachAI Guidance for Schools Toolkit	Vision, principles, policy review, sample guidance, stakeholder communication	Education authorities, district and school leaders	1–2	Developing or revising school and district guidance	Sample policies need local adaptation
Australian Framework for Generative AI in Schools	Teaching and learning; human and social well-being; transparency; fairness; accountability; privacy, security, safety	Policymakers, leaders, teachers, families, students	1–2	A national, principles-based governance baseline	Principles don’t resolve assignment-level questions
North Carolina DPI Generative AI Guidance	Leadership and vision; human capacity; curriculum, instruction, assessment; privacy; infrastructure	State, district, school, and instructional leaders	2–4	Connecting state guidance to district and classroom practice	Broad coverage needs local selection
CoSN/CGCS K–12 GenAI Maturity Tool	District self-assessment and planning across readiness dimensions	Superintendents, technology leaders, district teams	1	Diagnosing district maturity and setting priorities	Developed with corporate participation, including AWS
EdSAFE SAFE Benchmarks	Safety, accountability, fairness, transparency, efficacy	Procurement, policy, technology, governance leaders	1	Evaluating AI systems, vendors, risks, and safeguards	Doesn’t determine pedagogy
CIDDL Responsible AI Integration Guidance	Human oversight, transparency, law and policy, risk assessment, accessibility, instructional alignment	Districts, special-education and accessibility leaders	1–2	Governing high-impact and accessibility-sensitive use	Institutional, not assignment-facing

Governance, policy, and district-readiness frameworks

Assignment-permission and learning-design frameworks

The authors of the AI Assessment Scale (AIAS) emphasize that attaching a label to an unchanged assignment is inadequate: the task, rubric, checkpoints, and evidence have to be redesigned to match the level chosen. The University of Kentucky scale is unusual in classifying the student’s intellectual role instead of the tool’s presence.

That distinction earns its keep the first time you grade two essays that both disclose AI use. One student built the argument and ran the draft past the model for proofreading; the other wrote an outline and had the model produce the prose. A permission label that only records whether AI was used puts both in the same box. The Kentucky scale separates them by name, which is what you want in front of you at a parent conference.

Framework	Central variable	Structure	Proximity	Best use	Important limitation
AI Assessment Scale, Version 2 (AIAS)	Degree and nature of permitted AI use	No AI; AI Planning; AI Collaboration; Full AI; AI Exploration	4	Communicating expectations and redesigning assessments	Higher-education origin; one label can hide phase differences
North Carolina DPI / Braving Education 0-to-Infinity Scale	Classroom permission and integration	AI Free; AI Assisted; AI Enhanced; AI Empowered	4–5	Simple student-facing communication in K–12	Coarse categories need detailed teacher directions
NCTE stoplight approach	Simple permission category	AI-free; AI-supported; AI-driven	5	Rapid, understandable assignment labels	Too coarse for complex writing without phase-level explanation
University of Kentucky Student GAI Use Scale	The student’s intellectual role	Sole Author; Primary Creator; Conceptual Architect; Critical Collaborator; Project Manager	4	Clarifying ownership and division of labor in complex work	Higher-education origin; labels need secondary-level examples
WestEd Friction by Design	Which kinds of friction stay and which are removed	Cognitive ownership; productive struggle; social sense-making; activation energy; access	4–5	Deciding when AI removes required learning effort	Doesn’t itself state permission levels
TILT (Transparency in Learning and Teaching)	Transparency of purpose and expectations	Purpose; Task; Criteria	4–5	Making goals, process, and standards explicit to students	Not AI-specific; needs an added AI-role and evidence layer
Two-lane assessment	Whether performance is controlled or open	A controlled assurance lane; an open, AI-permitted lane	2–3	Designing a balanced course or program assessment system	Not granular enough for every task stage
Attempt → Coach → Verify → Transfer (ACVT)	Timing and scaffold sequence	Independent attempt; bounded AI support; verification; independent transfer	5	Designing a scaffolded AI interaction that preserves learning	Our own synthesis, not an externally validated framework

Assignment-permission and learning-design frameworks

WestEd’s five lenses reframe the permission question: good AI use removes friction that does not contribute to learning while preserving the friction that does. TILT predates generative AI and still holds up — add two fields to its purpose, task, and criteria, namely what role AI may play and what evidence of learning the student has to submit.

Which frameworks to stack, by role

A district leader working from governance principles alone still needs a readiness toolkit, a maturity diagnostic, and a procurement standard beside them. A classroom teacher needs a transparency structure, a friction lens, a permission scale, and something disciplinary. Pick by the decision you are actually making.

Your role or need	A workable stack
State or district leadership	UNICEF or Australian framework + TeachAI + CoSN maturity tool + EdSAFE or CIDDL
District curriculum office	UNESCO Students + OECD–European Commission AILit + Digital Promise or aiEDU
Teacher professional learning	UNESCO Teachers + TPACK-XK + SEE or aiEDU educator competencies
Secondary classroom teacher	TILT + WestEd Friction by Design + AIAS or Kentucky role scale + subject framework
ELA department	NCTE working framework + Kentucky role scale + TILT + secure writing samples
History or social studies department	AHA principles + OECD AILit or Digital Promise + Stanford CRAFT + permission and evidence framework
Interdisciplinary AI literacy or civics	UNESCO Students + OECD AILit + AI4K12 + CRAFT
Assessment office	AIAS + two-lane assessment + TILT + evidence-of-learning protocols
Technology procurement and safety	UNICEF + EdSAFE + Australian framework + CIDDL
A school wanting one simple first step	One traffic-light or four-level permission scale + one secure baseline + one AI-use disclosure format

Which frameworks to stack, by role

The social studies row is the thinnest one in the table, because no comparably mature, widely adopted, social-studies-specific generative-AI competency framework appears to have emerged. The American Historical Association supplies strong disciplinary principles and Stanford CRAFT supplies classroom activities, and neither is a complete grades 6–12 progression comparable to what UNESCO and the OECD offer for AI literacy in general. Departments filling that gap locally are working from the right instinct: use AI like someone in the discipline would. A historian interrogates a source’s provenance before quoting it, and so should a student handed a fluent paragraph by a model. The Four Modes of AI Assignments covers the assignment-permission layer of that work, and it is discipline-neutral, so the disciplinary progression is still missing.

Where the field agrees

Under the competing terminology most credible frameworks converge on ten points, and the tenth is the one worth putting first: knowing when not to use AI is part of AI literacy. Selective non-use is a competency, so a policy that only tells students how to use the tool has taught half of one.

The rest of the consensus:

Students and educators stay responsible for consequential decisions.
AI does not replace professional judgment or student accountability.
A tool earns its place by serving a learning goal.
Fluent output gets checked for accuracy, evidence, bias, omissions, and suitability.
Material AI use stays visible to the people it affects.
Privacy, equity, fairness, consent, access, and social effects are competencies in their own right.
Expectations change with a student’s age, knowledge, and experience.
A technically possible use can still be pedagogically, culturally, institutionally, or ethically unsuitable.
Policies get reviewed as the tools change under them.

One caution travels with all of it. Most of these frameworks are normative: they organize goals and decisions, and they do not establish that adopting them improves student achievement. Several of the assignment scales originated in higher education and need adapting for developmental level, parental expectations, compulsory schooling, and adolescent privacy. So a department can run a clean, well-labeled permission scale for a year and still have to look at what students can do in June to know whether it worked.

The Four Modes, ACVT, and the Five Tests

Three frameworks named across this site are our own work, and all three are practitioner syntheses drawn from the research rather than externally validated standards. The Five Tests are the checklist run before you authorize an activity at all; the Four Modes of AI Assignments is the permission scale you set once you have said yes, per assignment rather than per school; and Attempt → Coach → Verify → Transfer, in the table above, orders the events inside the lesson itself.

UNESCO and the OECD publish through international bodies with review processes behind them; ours came out of the research we did for these pages. Stack ours with something from the governance or competency families and they do the job they were built for. Two-lane assessment and the evidence question underneath it are worked through on our page about assessment design when AI is available. The rest of the cluster starts at our AI and K-12 hub.

Naming the proximity before the framework

Next time a policy question lands on you, settle its proximity before you go looking for a document. Governance, curriculum, and one assignment on Thursday are three different searches, and the tables above answer them in different rows.

Most schools already have an AI policy. Actually getting it into classrooms is the hard part. We help with that. Talk to us.