For Teacherss and Educators

Cognitive Sovereignty Checklist for Teacherss

About 20 minutes Last reviewed March 2026

When you use ChatGPT to plan lessons or Khanmigo to support students, the work feels faster and smarter. The danger is real. You might stop trusting your own read of a room, and students might submit polished work that hides shallow thinking. Both of those are harder to reverse than the time you saved.

Cognitive sovereignty insight for Teachers: a typographic card from Steve Raju

These are suggestions. Take what fits, leave the rest.

Download printable PDF

0 / 19 applicable

Tap once to check, again to mark N/A, again to reset.

Recognise When AI Is Replacing Your Classroom Judgement

Notice when you skip the live moment of confusionbeginner

When a student struggles, that struggle tells you something real. If you use AI to generate an instant explanation before you understand what the student actually got wrong, you lose information only you can gather by listening.

Record one lesson where you did not use AI planning, then compare it to one where you didintermediate

Teach the same topic two ways. With one, plan it yourself. With the other, build from an AI outline. Notice the difference in how you adapt mid-lesson, what you choose to skip, which tangents you follow. This shows you what AI planning takes from your practice.

Ask yourself what you would have chosen before the AI suggestion appearedbeginner

When Magic School AI or Diffit offers an activity for your Year 5 class, pause. What would you have designed? Did the AI suggest something better, or did you just accept it because it was there?

Track which decisions you make and which the AI makesintermediate

Over one week, note down: Did I choose the learning goal, or did AI? Did I pick the assessment task, or did AI? Did I decide who needed extra support, or did AI flag it? You should own most of these calls.

Refuse one AI suggestion per week on purposebeginner

When Gemini or ChatGPT offers an idea, say no at least once a week. Practise trusting your own judgement over the tool. This keeps your decision-making muscle strong.

Identify the moment you stopped assessing and started sortingadvanced

Assessment means you form a picture of what a student understands and where they are stuck. Sorting means you run their work through a checklist or let AI mark it as done or not done. These are not the same. Know which one you are doing.

Design Assessment That Forces Real Understanding Into View

Build assessment tasks that AI cannot do for studentsintermediate

If the task is to summarise or explain a concept, a student can ask AI to do it. Instead, ask students to apply the concept to a problem they choose, explain why it matters to them, or argue against it. These tasks show what they actually know.

Require students to show their working and their thinkingbeginner

Do not ask for an answer. Ask for the thinking path. How did you decide? What made you reject the other option? What would change your mind? A student might copy an AI answer, but they cannot easily copy their own reasoning.

Conduct one oral assessment per term with each studentintermediate

Ask them to explain their work in a real conversation. You can ask follow-up questions, watch their face, hear them think out loud. No AI can impersonate this. You will know what they understand.

Create assessment criteria that reward revision and change of mindbeginner

Tell students you want to see their first attempt, their second attempt, and their thinking about what changed. This makes it harder to hide behind a polished AI output. It also teaches that learning is not finished after one draft.

Ask students to assess AI work against your criteriaintermediate

Show them a piece of writing or a lesson plan made by AI. Have them mark it using your assessment rubric. This teaches them to spot where AI sounds good but misses the point. It also teaches them your standards through comparison.

Use assessment to check for transfer, not just recalladvanced

Do not assess whether students remember what you taught. Assess whether they can use it in a new situation. A student who can copy AI explanations cannot necessarily do this. Transfer shows real understanding.

Mark work by hand at least once per unitbeginner

Read a stack of student work yourself, without AI help. Notice what you learn from the handwriting, the crossings-out, the margin notes. Notice which errors appear in multiple students. This information guides your next lesson better than an AI report.

Reclaim the Craft of Teaching From AI Efficiency

Spend thirty minutes planning each lesson without opening an AI toolbeginner

Think on paper or screen first. What do you want students to be able to do? What are they getting wrong now? Only then ask an AI tool to help with resources. Your own thinking comes first.

Protect the adaptations that only happen liveintermediate

A lesson plan from AI is a skeleton. You add life in real time: you notice that three students look lost and spend five more minutes on one idea, you spot a connection and chase it, you drop a planned activity because momentum is better elsewhere. These moves make you a teacher, not a content deliverer.

Keep a notebook of analogies and examples that workbeginner

When you find or invent an explanation that helps students understand something hard, write it down. This becomes your resource library. You own it. You know why each one works with your students. Use this before asking AI to generate an analogy.

Run an experiment: teach one topic AI-free, one topic with AI supportadvanced

Compare the two. Which felt more like teaching? Which gave you better information about what students learned? Which lesson did students remember more about later? Use what you learn to set your own boundaries with AI.

Tell your students the truth about when and how you use AIbeginner

Say: I used AI to generate quiz questions, but I wrote the learning goals myself. I asked ChatGPT for ideas about this topic, but I chose which idea to teach you. Being honest about your choices teaches them to be honest about theirs.

Resist the pressure to personalise every activity with AIintermediate

Whole-class teaching, shared struggle, and common learning matter. Not every student needs a bespoke AI-generated worksheet. Sometimes the strength of teaching is teaching the whole class the same thing and watching different minds solve it.

Five things worth remembering

When an AI tool offers to mark work or generate feedback, sit with one piece of student work and mark it yourself first. Compare your feedback to what the AI gave. You will see what matters that the tool missed.
Ask your students: Show me one thing an AI tool helped you with this week, and one thing you figured out yourself. This simple question makes them notice the difference between help and outsourcing their thinking.
Set a rule: You must understand a student's confusion before you can explain it away. If you skip this step and jump to an AI explanation, you have stopped teaching.
Use AI for the work you hate, not the work you love. Use it for marking 200 multiple choice responses, not for designing your Year 9 poetry unit. Protect the parts of teaching that are craft.
Every term, teach one lesson the way you would have taught it ten years ago, with no AI. Notice what you miss. Notice what is better. Let the comparison guide where AI belongs in your practice.

Prompt Pack

Paste any of these into Claude or ChatGPT to pressure-test your own judgment. They work best when you respond honestly before reading the AI reply.

Test your own lesson design thinking

I am planning a lesson on [topic] for [year group / age]. Before I look at any AI-generated plans, ask me questions about what my specific students need, where they typically get stuck, and what I am trying to achieve. Then I will compare my thinking to an AI plan.

Audit your feedback habits

I have been using AI to help write feedback on student work. Ask me questions that help me examine whether this feedback genuinely reflects my knowledge of these students or whether it has become generic. What have I stopped noticing since I started using AI for this?

Design an AI-proof assignment

I need to redesign an assignment that students are currently completing using AI without learning anything. The topic is [describe]. Help me redesign it so that the learning process, not just the output, is what gets assessed.

Rebuild your reading of student understanding

Describe a student response or piece of work and ask me what I think it reveals about their understanding, before I see any AI analysis. Help me practice making my own diagnostic judgments about student learning.

Challenge your reliance on AI curriculum resources

I have been using AI-generated materials in my classroom. Ask me how much of my current curriculum planning reflects genuine pedagogical decisions versus convenient AI defaults. What have I stopped thinking about carefully since I started using these tools?

Reading List

Five books that give this topic the depth it deserves. Each one is genuinely worth reading, not just citing.

Make It Stick

Peter Brown, Henry Roediger and Mark McDaniel

The science of how learning actually works, retrieval, spacing, interleaving, desirable difficulty. Every principle here is undermined by students using AI to skip the struggle.

Why Don't Students Like School?

Daniel Willingham

A cognitive scientist's account of how memory and thinking work in learning. The implications for AI-assisted education are significant and under-discussed.

The Shallows

Nicholas Carr

What persistent use of tools that do the thinking for us does to the capacity for sustained reading and deep thought. The foundation of education.

Range

David Epstein

The case for breadth and transferable thinking over narrow specialisation, a direct counter to AI systems that optimise for narrow task performance.

Cognitive Sovereignty

Steve Raju

A framework for protecting independent thought that is directly applicable to what you want your students to develop. And what AI is quietly dismantling.

Questions to ask yourself

Use these before your next AI-assisted decision. Honest answers are more useful than comfortable ones.

When did I last design a lesson from scratch based on my knowledge of these specific students, without starting from an AI-generated plan?
Is my feedback on student work genuinely individual, or has it become generic since I started using AI assistance?
Am I building my pedagogical judgment over time, or delegating decisions I should be making myself?
What do I know about my students that no AI curriculum tool has access to. And am I acting on it?
Are my students developing the capacity to think through difficulty, or am I enabling them to avoid it?

Common questions

Should teachers use AI tools in the classroom?

AI tools can reduce the administrative burden of teaching, generating differentiated resources, writing feedback on drafts, and helping plan lesson sequences. The question is whether using AI for these tasks comes at the cost of the pedagogical thinking that makes you an effective teacher. Lesson design built on your knowledge of your specific students is different from AI-generated lesson design built on generic learner profiles.

How can teachers tell if students are using AI on their work?

Detection tools exist but are unreliable and produce false positives. More effective approaches involve assignment design that makes AI-generated responses obviously inadequate: in-class discussion of submitted work, oral components, assignments that require personal experience, and iterative drafts. These also happen to be better learning designs regardless of AI.

What are the risks of AI for education?

The deepest risk is not cheating. It is the erosion of the productive struggle that builds genuine competence. When students bypass the difficulty of formulating their own ideas, constructing arguments, and working through confusion, they do not develop the cognitive capability that education is meant to build. AI makes the output easy and hollows out the process.

Will AI replace teachers?

AI can personalise content delivery and provide instant feedback at scale, which are parts of teaching. But the relational work of education, reading where a student is emotionally, knowing when to challenge and when to support, building the trust that makes learning possible, is not a content delivery problem. Those skills are not at risk of automation; they are at risk of neglect if teachers over-delegate to AI.

How should teachers respond to students using AI?

Treat it as a curriculum question rather than a discipline problem. Help students understand what AI does to their learning process, that the struggle is not an obstacle to learning, it is the mechanism of it. Build assignments that require students to demonstrate understanding through dialogue, reflection, and application rather than just finished text.