For Teacherss and Educators

Protecting Your Judgement: A Teachers's Guide to Using AI Without Losing What Makes Teaching Work

AI tools like ChatGPT and Khanmigo can generate lesson plans and answer student questions in seconds. But when a tool handles the thinking work, you lose the moments where real teaching happens: noticing confusion in a child's face, finding the example that finally clicks, deciding to abandon your plan because the room needs something different. Protecting your judgement means knowing which parts of teaching AI should never touch.

These are suggestions. Your situation will differ. Use what is useful.

Download printable PDF

Know What You Cannot Delegate to AI

Your job includes things only you can do: reading a room's mood, hearing the question behind the question, choosing whether to push or reassure a particular child. When you hand these decisions to AI, you stop being a teacher and start being a content delivery system. Use AI for generating multiple-choice distractors, summarising reading material, or creating quiz templates. Do not use it to decide whether a student understands something, to write their end-of-unit feedback, or to plan how you will introduce a difficult concept to your specific Year 7 class.

›Use Magic School AI to generate practice problems, but write your own explanations based on what you know confuses your students
›Let ChatGPT draft ten different starter activities, then choose the one that matches your class's readiness and energy today
›Use Diffit to adapt texts to different reading levels, but make the final call on which version each student gets based on your observations

Design Assessment So AI Cannot Hide Student Gaps

The real danger is not that students use AI to write essays. The danger is that you cannot tell whether they understand. If your assessment is multiple choice or a one-paragraph summary generated from a prompt, AI-generated work looks identical to student work. Change how you assess so understanding becomes visible. Ask students to explain their thinking step by step. Require them to apply ideas to new situations. Use low-stakes quizzes where speed matters and copying is obvious. When you assess this way, a student who has not understood cannot hide behind a tool.

›Replace end-of-unit essays with live verbal explanations recorded on your phone where you ask follow-up questions
›Use short diagnostic quizzes given in class on paper, where you watch how students approach problems and notice where they get stuck
›Ask students to annotate source material explaining their own thinking, so you see their reasoning not just their conclusion

Use AI for Lesson Planning, But Protect Your Craft

Khanmigo and Magic School AI can generate lesson outlines fast. This is genuinely useful when you are planning five lessons before Tuesday. What matters is that you do not outsource the decisions that define your teaching. AI can suggest a learning objective, but you decide if it matches what your students actually need to learn. AI can propose activities, but you choose which ones fit your classroom, your students' pace, and your professional judgment about what teaching looks like in your context. The tool creates options. You decide.

›Ask ChatGPT to generate five different ways to explain photosynthesis at GCSE level, then pick the one that matches how you actually teach it
›Use Diffit to create differentiated worksheets for a lesson you have planned, not to decide what the lesson should be
›Generate a five-lesson unit outline with ChatGPT, then spend 30 minutes rewriting each lesson so it reflects your actual teaching moves and your knowledge of your class

Make Student Independence Real, Not Apparent

Students now have access to tools that will do their thinking for them. This means you must design work that requires genuine independent thinking or makes shortcuts obvious. A student who types a question into ChatGPT and submits the answer has not done the work. A student who uses ChatGPT as a research assistant, checks its accuracy against real sources, and writes her own argument based on what she learns might have done something valuable. The difference is whether the student had to do the cognitive work. Set assignments where short-cutting does not work. Require students to show their process, not just their answer. Use live problem solving in class where you watch understanding happen.

›Ask students to solve a problem in three different ways, proving they understand the concept not just the answer
›Require rough notes and rough drafts with timestamps, so you see the thinking process over weeks not just the final product
›Use Khanmigo's built-in structure to have it quiz students after they claim to understand, making gaps obvious to both student and teacher

Reclaim the Parts of Teaching That Matter Most

If AI handles content delivery and routine planning, what actually becomes scarce and valuable? Your presence. Your attention. Your ability to see a struggling child and adjust. The fact that you remember what this class found difficult last week. Your judgment about what a child is ready for. These are not things AI will ever replace, but they are easy to lose if you let AI absorb all your planning and marking time. Protect them deliberately. Build class time that is genuinely interactive. Keep some marking work that lets you write real feedback tied to what you know about each student. Make space to think about your teaching, not just deliver it.

›Use AI to mark routine work and create summaries, so you spend marking time writing personal feedback on extended pieces where you see and respond to individual thinking
›Block one day per week where you plan no lessons in advance, leaving space to respond to what actually happened in your classroom
›Keep one regular interaction with each student per week that is not AI mediated: a question during independent work, a comment on something they said, a check-in about their confidence

Key principles

1.If the task involves reading a human in front of you and deciding what that person needs, do it yourself.
2.Assessment that cannot distinguish between student thinking and AI output is broken assessment and must be redesigned.
3.AI is useful for generating options and raw material, but your professional judgment about what is right for your class is irreplaceable.
4.Student independence means the cognitive work happened in the student's head, not in how polished the final product looks.
5.Teaching happens in the unplanned moments and the noticing and the adjusting, which is why you must protect time and attention for these things deliberately.

Key reminders

When you feel tempted to use an AI tool for something, ask first: could I do this without thinking about my specific students? If yes, the tool is probably safe to use. If the answer is no, you should probably do it yourself.
Set a rule for yourself: you will use ChatGPT or Gemini to generate five options for something, and you will rewrite all five before using any of them. This keeps you in the driver's seat.
Create a checklist for every assignment you set: could a student submit pure AI output and get full marks? If yes, rewrite the assignment so the answer is no.
Keep a log of the moments in your week when you made a professional judgement that a tool could not have made. Do this for a month. You will see clearly what teaching actually is.
Talk to your students directly about AI. Tell them which tools you use and which thinking they need to do. Most students will take you seriously if you explain why genuine understanding matters more than speed.