The situation you are actually in
Your students have access to these tools on their phones, at home, at two in the morning, with nobody watching and nobody explaining anything. You can decide whether AI appears in your classroom. You cannot decide whether it appears in their lives, and the only guidance they get otherwise comes from friends, from social media, and from the tool itself, which has no ability to tell them when it is wrong.
So the question is not whether to allow it. It is whether they hear anything sensible about it from an adult.
Three things are worth teaching, and they hold from primary through further education with the language changed: disclose what you used, verify what it tells you, and use it to understand rather than to produce.
In plain English
- Disclosure:
- Saying what you used and how. Not a confession, just part of showing your working.
- Verification:
- Checking a claim against a source you trust before you rely on it.
- Hallucination:
- Invented content stated confidently. Fake quotations, fake sources, fake dates, fake page numbers.
- Cognitive offloading:
- Letting something else do the thinking. Fine for arithmetic you have mastered, costly for skills you are still building.
Disclosure, and why it has to be safe
Disclosure only works if telling the truth is not punished. If every admission of AI use triggers suspicion, students stop admitting anything, and you lose the only reliable window you had.
So say clearly what is fine, and make disclosure ordinary. A line at the end of a piece of work: what I used it for, what it gave me, what I changed. Treat it the way you treat a bibliography, as evidence of good practice rather than an admission of weakness.
For younger children this is simpler. Did the computer help you? What did it help with? Did you check whether it was right? Children are usually delighted to explain.
Write me a three line disclosure box for students to complete
at the end of a piece of work.
Year group: [ ]
Subject: [ ]
Requirements: three short prompts asking what they used, what it
produced, and what they changed or rejected. Written so a student
does not feel accused by filling it in. Under 40 words in total.
Give me two versions, one for younger students and one for older.
Verification, taught by demonstration
Telling students that AI makes things up does not work. Showing them does, and it takes ten minutes.
Ask the tool something in your subject where you know the answer cold and the detail is obscure. Ask it for sources, then check them together in front of the class. Some will be real. Some will be plausible and entirely invented, with convincing authors and page numbers. Watching a confident, well-formatted, fabricated reference collapse under one search does more than any warning you could give.
Do it again with something the class thinks it knows. Ask the tool to explain a topic they have just studied and get them to mark it. Marking the machine teaches the material better than reading it would have done.
Write an explanation of [topic] for [year group] students.
Include at least two errors: one factual, one where the reasoning
looks right but is not. Do not tell me where they are.
Keep it to 250 words and make it sound like a normal confident
explanation, not a trick.
Then, separately and clearly labelled at the bottom, give me the
answer key explaining each error and why students might miss it.
Checkpoint
Three things to teach: disclose what you used, verify what it tells you, and use it to understand rather than to produce. Disclosure only works if honesty is safe.
Understand, do not produce
This is the distinction that actually matters, and students grasp it quickly when it is put plainly.
Using it to understand looks like: explain this to me differently, I am stuck on step three, quiz me on this chapter, what am I not getting here, give me three more practice questions like this one. The student is doing the thinking. The tool is a patient explainer that never sighs.
Using it to produce looks like: write my essay, do my homework, summarise this book I did not read. The student has skipped the part where the learning happens and handed in a plausible object.
Say it to them straight. It is not about rules, it is about what happens to you. If a tool writes your essays for two years, you cannot write. That is not a punishment, it is just how skills work, and you will find out at exactly the wrong moment.
โ Weak prompt
Prompt
Write me a 500 word essay on why the Treaty of Versailles caused the Second World War.
Output
A competent, structured, entirely usable essay produced in fifteen seconds.
The student now owns a document and has learned nothing. They cannot defend a single sentence of it, and if this becomes a habit the ability to build an argument never develops.
โ Good prompt
Prompt
I am writing about whether the Treaty of Versailles caused the Second World War. Here is my argument in three sentences. Ask me five questions that would expose the weakest part of it. Do not write anything for me and do not tell me the answers.
Output
Five uncomfortable questions about causation, timing and the evidence being leaned on.
The student still writes every word, but the argument gets stress-tested first. This is the use worth teaching, and it produces better essays than the shortcut does.
Age-appropriate framing
Primary. Keep it concrete and short. The computer can be helpful but it is sometimes wrong, and it never says sorry or looks embarrassed. Always tell a grown-up what it helped with. Never tell it anything private about yourself, your family or your friends.
Lower secondary. Introduce verification properly and make disclosure routine. Focus hard on the difference between getting help understanding and getting something produced for you, because this is the age at which the habit forms.
Upper secondary and further education. Be honest about the stakes. Exam boards and universities have rules and they matter. So does the fact that they are entering workplaces where these tools are normal and the people who do well are the ones who can judge the output rather than accept it. Framing it as a skill they will need lands better than framing it as a temptation.
Two rules that apply at every age. Never put personal information about yourself or anyone else into these tools. And never treat them as a source of advice about your health, your safety or anything you are worried about. Those conversations are for people.
Plan a 30 minute lesson that teaches [year group] students to check
what an AI tool tells them.
It must include:
- one live demonstration I run at the front
- one activity students do in pairs, no devices needed
- three specific things to check that they can remember later
- a short discussion question about why the tool sounds confident
even when it is wrong
Do not moralise and do not include a list of rules.
The point is that they see it happen, not that they are told.
The most useful sentence you can give a class: it is not lying to you, it is guessing what words come next, and it has no idea when it is wrong. That single idea explains almost every failure they will meet.
๐ Quiz
Question 1 of 4Why is teaching responsible AI use worth doing even if you do not use AI in your own teaching?