Implementation Practical Playbook 20 to 30 minute material review

Keep, Revise or Reject? How to Review AI-Generated Lesson Materials

An AI tool can produce a polished lesson plan, worksheet or quiz quickly. This seven-part check helps teachers decide what to keep, what to fix and what should never reach pupils.

In this playbook9 parts
  1. Prepare for the session
  2. 01 Does the material preserve the curriculum purpose?
  3. 02 Are the content and answer paths correct?
  4. 03 Does the activity require the intended thinking?
  5. 04 Does the support respond to real learner barriers?
  6. 05 Is representation fair and relevant?
  7. 06 Can intended pupils perceive, navigate and respond?
  8. 07 Has a teacher adapted and rehearsed the final version?
  9. Completion check

The worksheet looks finished. The instructions are tidy, the answer key is attached and a box labelled differentiation sits at the end. Then the first worked example is wrong, or the objective asks pupils to explain while every task asks them to copy. Polished material is not necessarily ready material.

Use the fictional example of a lower-secondary science pack on changes of state as you work. It contains a misleading particle explanation, a colour-only diagram, mostly recall tasks, vague support, weak extension and 50 minutes of activity for a 40-minute lesson.

Keep a material only after a teacher has verified its curriculum purpose, subject accuracy, cognitive demand, learner fit, representation, accessibility and classroom feasibility. Revise visible flaws that can be corrected and checked again. Reject a critical problem that remains unsafe, inaccurate, unverifiable or misaligned.

This is a review of one material, not a rating of the tool. The same tool may produce a useful starting point for one task and an unusable result for another.

01

Does the material preserve the curriculum purpose?

The activity asks pupils to perform the knowledge, thinking or disciplinary process named in the objective and fits its place in the teaching sequence.

Write the objective in your own words, then inspect what pupils will actually do. Watch for goal drift, where the topic appears correct but the task changes the intended learning. In the science example, matching labels does not demonstrate an explanation of changes of state.

Research on curricular alignment in AI-generated biology lesson plans shows why surface alignment is not enough when the form of learning changes.

Ask to see
The exact objective, curriculum expectation, prior knowledge, pupil activity and assessment response.
Pause if
The topic matches but the task substitutes recall, copying or decoration for the intended explanation, reasoning or practice.
Write down
The intended learning and whether the task and assessment preserve it.
02

Are the content and answer paths correct?

A teacher with sufficient subject knowledge has independently checked every consequential fact, instruction, worked example, answer and alternative solution.

Work each answer yourself and check the reasoning path, not only the final response. Fluent wording can make a plausible error difficult to notice, and asking the same AI system to approve its own output is not independent verification.

A study of preservice mathematics teachers evaluating a ChatGPT solution illustrates how an incorrect solution can be mistaken for a different approach.

Ask to see
Facts, calculations, quotations, technical vocabulary, instructions, worked examples, answer keys, rubrics and valid alternative solutions.
Pause if
A material answer is wrong, cannot be verified or depends on reasoning that the teacher has not checked.
Write down
The checked answer path, the corrections made and any unresolved subject question.
03

Does the activity require the intended thinking?

Pupils must generate, connect, predict, justify or respond to reasoning when the objective requires more than recognition or routine practice.

Look past labels such as discussion, inquiry or project. Watch for cognitive substitution, where the material performs the important thinking and leaves pupils to copy, select or decorate the result.

The ICAP framework offers a useful lens for distinguishing observable forms of engagement without turning them into a universal score.

Ask to see
The observable pupil action in each task and its connection to the stated objective.
Pause if
The material supplies the explanation or reasoning that pupils were meant to produce.
Write down
The level of pupil thinking required and the task change needed to restore it.
04

Does the support respond to real learner barriers?

Each adjustment names a likely barrier or misconception and helps the pupil approach the same learning goal without silently lowering it.

Check prior knowledge, vocabulary, reading load, working-memory demands, modelling and rehearsal. Generic easier work, extra time or more questions are learner-blind when they do not address why a pupil may struggle.

A study of AI-generated lesson-plan alignment with UDL and UDT found that teacher-initiated modification remained necessary in its evaluated context.

Ask to see
Likely misconceptions, language demands, access barriers, modelling needs and a specific adjustment for each material barrier.
Pause if
Support changes the learning goal, relies on vague ability labels or adds work without addressing a defined barrier.
Write down
The barrier, the adjustment and how the intended learning remains intact.
05

Is representation fair and relevant?

Names, roles, settings and examples avoid stereotypes, distribute agency fairly and do not treat one family, language or school experience as universal.

A material does not need to represent every identity in every example. It does need to avoid tokenism, misleading generalisations and patterns that assign authority, competence or difficulty unevenly.

The Department for Education's guidance on generative AI in education keeps responsibility for checking inaccurate, biased or out-of-context output with education professionals.

Ask to see
Who has agency, expertise and voice, which experiences are treated as normal and whether cultural knowledge creates an irrelevant barrier.
Pause if
Visible diversity masks unequal expectations, stereotypes or missing perspectives that distort the task.
Write down
What passes, what needs changing and which perspective or context is missing.
06

Can intended pupils perceive, navigate and respond?

Essential information is available in more than one appropriate form, the reading order and instructions are clear, and the response mode remains accessible without changing the construct being assessed.

Accessibility is not the same as simplified language. Check text alternatives, colour use, headings, vocabulary, symbols, device support and whether every intended pupil can take part in the required response.

Use WCAG 2.2 for digital accessibility checks while keeping pedagogical quality as a separate judgement.

Ask to see
Text alternatives, non-colour cues, logical headings and reading order, explained symbols and an accessible response route.
Pause if
Essential information is visual only, colour carries meaning alone or the required response excludes an intended pupil without a pedagogical reason.
Write down
The access barrier, the correction and how the learning purpose is preserved.
07

Has a teacher adapted and rehearsed the final version?

The reviewed version fits the actual lesson time, resources, terminology, grouping, devices and likely pupil responses, with critical instructions and answer paths rehearsed.

Check the complete adapted material again. Classroom infeasibility often hides in transitions, equipment, printing, technology, explanations and timing that looked manageable on the page.

The Shiksha Copilot study provides one example of layered curriculum grounding, curator review, translation review and teacher customisation.

Ask to see
The final local version, realistic timings, required resources, grouping, key explanations, likely questions and a contingency for failure.
Pause if
The plan cannot run in the available setting, depends on unchecked instructions or answers, or has no workable contingency.
Write down
The final changes, the rehearsal check and the person responsible for any remaining action.

Turn individual checking into a shared routine

If your school is moving from individual AI experiments to consistent teaching practice, TopSchool can help your team review the workflow, responsibilities and safeguards around classroom use.

Review your AI teaching workflow
TopSchool logo
About the TopSchool Team

TopSchool Team

Education, technology and implementation team

Some of the most useful school AI guidance is shaped by more than one discipline. The TopSchool Team brings together perspectives from education, technology, school partnerships, product and implementation to explain what schools need to consider next.

School AI implementationProduct guidanceCross-functional practice
View profile and articles