No preparation, no slides to make, no software to install. One link, four sections, and a set of discussion questions that work whether or not you know anything about AI yourself.
By the end, students can explain what an AI model is actually doing, and can recognise a confidently wrong answer rather than accepting it.
That is the whole objective, and it is deliberately narrow. A lesson that lists ten AI tools is out of date in a term. A lesson that explains next-word prediction is still true in five years.
Because you asked. It continues text, and your question was a shape that needed filling. It was not lying — lying requires knowing the truth. Follow-up: ask when this would matter most. Answer: when they know least about the topic, which is exactly when they were relying on it.
You could not, from the answer alone — it is written in exactly the same handwriting as a correct answer. You have to check the source. Follow-up: which things does it invent most and students check least? Names, dates, numbers, citations.
No — and this is the important correction. It is excellent at anything made of language: summarising, rewriting, translating, finding. It is weak at arithmetic, at facts about the world, and at anything somebody must answer for. Follow-up: ask for one example of each from their own homework.
Turn this back to the class rather than answering it. Ask them to draw the line themselves, then compare with your school's rule. The distinction most classes reach on their own: using it to understand something is different from using it to produce something you then claim you wrote.
Ten minutes with Level 1 will tell you more about whether this suits your class than anything on this page.
Open the activity