Quote it or it did not happen: teaching kids to demand a source
The single most useful sentence a child can learn to type to an AI is "show me the exact line." Here is why that question matters and how to teach a child to always ask it.
We planted one fake fact into a set of class notes before a session - "our school mascot is a penguin," which it is not - and asked the AI a question that touched on it. It repeated the penguin fact back with total confidence, because it had been told to answer from the document, and the document said so. The children found this hilarious. They also immediately understood something most adults take years to internalise about AI: an answer is only as good as what it was allowed to read.
The fix for that entire category of problem is one habit, repeated relentlessly: ask it to quote the exact line, and then go and check the line is really there.
How do I know if an AI answer came from my file?
Ask it directly. "Which exact sentence in the document says that? Copy it out." A genuinely source-grounded answer will produce an exact quotation you can locate on the actual page. An answer with no real source behind it will either refuse, go vague, or - worse - invent a plausible-sounding quotation that is not really in the document at all.
"Class_notes_volcanoes.docx, line 14: 'the model must use baking soda and vinegar, not yeast.'" Locate line 14. It says that.
A confident quotation that sounds exactly right - but line 14 says something else entirely, or the document does not contain that sentence at all.
Quote it or it did not happen
That phrase is the entire lesson, and it is deliberately blunt. A source you cannot check is not a source - it is just a claim wearing a source's clothes. Teaching a child to demand the exact sentence, every time, turns a habit of trust into a habit of verification without making them cynical about the tool as a whole.
The honest limit
This habit catches a fabricated quotation. It does not catch an AI faithfully quoting a document that was itself wrong, out of date, or - as in the exercise above - deliberately seeded with something false. Verification checks whether the AI is telling the truth about its source. It does not check whether the source itself is true, and that second check still belongs to the reader. This is the last article in the knowledge level; from here, level 5 is about turning a checked, trusted process like this one into something a child can save and reuse by name.
Questions we get asked
How do I know if an AI answer came from my file?
Ask it to quote the exact sentence from the document that the answer is based on, then locate that sentence in the real file yourself. A genuinely grounded answer produces a real, findable quotation. An answer with no real source behind it will go vague, refuse, or invent a plausible-sounding quotation that isn't actually in the document.
Can an AI make up a fake quote and present it as a real source?
Yes, this happens, and it is why quoting alone is not enough - the quote has to be checked against the real document. Treat any citation as a claim to verify, not proof in itself, until you have located the exact line it points to.
What is the best habit for teaching kids to check AI sources?
Teach the phrase "quote it or it did not happen" - insist the AI point to an exact sentence for any factual claim from a document, and make checking that sentence a non-negotiable last step, every time, not just when something sounds suspicious.