Be the person who checks.
A cheerful sentence can sneak in a guess. Can you catch it?
Pause, check, ask
Catch a pretend mistake and practice checking a confident answer with an adult and a dependable source.
Read the video out loud
An AI answer can sound very sure and still be wrong. Here is a pretend answer: a triangle has four sides.
Does that match what you know? Let's check a trusted book with an adult. A triangle has three sides.
We found the mistake! For harder questions, ask a trusted adult to help find a dependable source.
Do not use a chatbot to decide what medicine to take or what to do in an emergency.
Pause, check, and ask for help. That is smart learning. Stay curious, and keep checking what matters.
Fact, guess, or mismatch?
Read the original note. Decide whether each claim matches it, adds an unknown detail, or disagrees with it.
Hear Patch’s short reminder. Your grown-up can read the full cards below.
Spoken words: Read the picnic note with your grown-up. Does each answer match, add a guess, or disagree?
Play without the buttons
Read the picnic note above. Does it support “The picnic is Saturday”? Yes. Does it support “It starts at two o’clock”? No—the time is undecided. Does “The picnic is Sunday” match? No—it contradicts Saturday.
Point to the exact line that helps you decide. If the note does not tell you, leave the detail unknown.
Your choices stay in this tab. There is no score to save, no ranking, and no account. Start again resets this game; closing the page clears its game state.
Sounding sure is not the same as being right.
AI can produce a sentence that sounds sensible but adds something nobody supplied. Compare the answer with the original information. Ask a trusted grown-up when a claim matters or is hard to check.
These answer cards were written as teaching examples. They are not real responses from our hosted model.
Go deeper: what counts as evidence?
Try “Which line supports that?” For a real-world claim, look for the original source, its date, and what it actually measured. A chatbot’s confidence or another chatbot agreeing is not enough.
A useful answer can say, “The information does not tell us.” That leaves room to find out.