Pupils will use AI throughout their lives. The most valuable thing schools can teach isn't how to prompt it — it's how to judge whether what it produces is any good. Here's how to build that at KS4.
Why this belongs in the curriculum now
Pupils are already getting answers from AI, and those answers are fluent, confident and sometimes wrong. Traditional source evaluation taught pupils to check who wrote something and why — but AI output has no author, no visible motive and no citations, so the old questions don't transfer neatly. This is a genuinely new literacy, and it maps well onto existing critical thinking work in English, history, science and RE.
The core idea to teach: fluency isn't accuracy
The single most important lesson is that AI text sounds authoritative regardless of whether it's correct. Pupils are used to judging credibility partly by how something is written — AI breaks that heuristic completely. Everything else follows from understanding this.
Activities that work
1. Spot the error
Give pupils AI output containing planted mistakes and ask them to find them, using their subject knowledge and other sources.
Prompt — generating material for this
Write a confident 200-word explanation of [topic] at GCSE level containing three factual errors and one significant omission. Don't indicate where they are. List them separately for me.
2. Compare AI against a reliable source
Same question to AI and to a textbook or authoritative site. Where do they differ? Which do you trust, and why?
3. Hunt the fake citation
Ask AI for sources on a topic, then have pupils try to verify each. The experience of discovering that a plausible-looking reference simply doesn't exist is far more persuasive than being told it happens.
4. Same question, different framing
Ask AI the same question phrased neutrally and then leadingly. Pupils see how the framing shapes the answer — a direct route into bias.
5. What's missing?
Ask AI for arguments on a contested topic, then have pupils identify whose perspective is absent. Excellent in history, RE and geography.
The questions pupils should internalise: How would I check this? What would make this wrong? Whose view is missing? Does this cite anything real? Would I recognise an error here if there were one? That last question is the sharpest — it teaches pupils that AI is least safe precisely where they know least.
Where it fits
It doesn't need a new subject slot. English can build it into source and language analysis, history into source evaluation, science into experimental evidence and claims, RE and PSHE into bias and perspective, and computing into how the technology actually works. A coordinated approach across a few departments is more effective than a one-off assembly.
Model it honestly yourself
The most powerful teaching here is showing pupils your own use — including where AI got something wrong and how you spotted it. Pupils who see a teacher checking AI output learn far more than pupils who are told to. See designing tasks around AI and the honest position on detection.
Want to build AI literacy across your school?
We help schools teach pupils to use AI thoughtfully, not just safely. AskColin gives schools low-cost monthly support with one-to-one help whenever you need it — so when you're stuck on a real job at 8am, there's someone to ask. Practical, jargon-free, built around your department.
Get one-to-one AI support for your schoolFrequently asked questions
How do you teach pupils to evaluate AI output?
Through practical activities: spotting planted errors in AI text, comparing AI answers against reliable sources, trying to verify AI-generated citations, asking the same question with neutral and leading framing to reveal bias, and identifying whose perspective is missing.
Why is AI evaluation a new literacy?
Traditional source evaluation asks who wrote something and why, but AI output has no author, no visible motive and no citations, so those questions don't transfer. Crucially, AI text sounds authoritative regardless of accuracy, which breaks the heuristic pupils normally use to judge credibility.
Where does AI critical thinking fit in the curriculum?
It doesn't need a new slot — English can build it into language and source analysis, history into source evaluation, science into evidence and claims, RE and PSHE into bias and perspective, and computing into how the technology works. A coordinated approach beats a one-off assembly.
What questions should pupils ask about AI output?
How would I check this? What would make this wrong? Whose view is missing? Does this cite anything real? And most sharply: would I recognise an error here if there were one — which teaches that AI is least safe precisely where pupils know least.