Uma casa em 3D, um pequeno filme ou um jogo de tabuleiro: comecem com um projeto de IA em família grátis. Conhecer a adesão →
← Projetos de IA em família

Para pais e filhos fazerem juntos · Projeto gratuito

🔍 A IA está certa? Seja o pequeno professor dela

Uma folha de verificação em papel, um erro da IA encontrado e uma correção confirmada em um livro real.

Cerca de 15 minutos · Dos 6–12 anos

As instruções estão disponíveis em inglês e chinês. Esta página mostra a versão inglesa.

Free Tonight · 10–15 Minutes (Optional Deeper Book Check up to 20 Min)

Start Right Away: Read the Zero-Account Card or Run Live in Adult ChatGPT

Zero-Account Instant Card · Authored Teaching Example (Not Live AI Output)Ages 6–8 · Penguins

Read these 3 sentences aloud — let your child guess the sneaky mistake first:

  1. Penguins are flightless birds that use their stiff flippers to “fly” underwater.
  2. Most wild penguins live on the icy Arctic sea ice right next to polar bears.
  3. Penguins have a thick layer of blubber and tightly packed feathers to stay warm.
Optional Live Round · Adult-Operated ChatGPT PromptOpenAI 13+ rule: adult operates keyboard · no child account or PII

Prefer a live AI round on this topic (or swap in dinosaurs, volcanoes, or soccer)? Paste into your own adult ChatGPT account while your child listens and grades on paper—never enter your child’s name, school, or photo.

Write 3 short sentences for a family fact-checking game about penguins for a 6-8-year-old. Include 2 accurate scientific facts and 1 subtle factual mistake that sounds believable. Do not reveal which sentence is the mistake yet.
Free ChatGPT tier works; zero paid AI subscription needed.

Tonight your child gets to grade the AI. Three short sentences, one of them confidently wrong: your child predicts, circles the sneaky mistake, and checks it in a real book. It takes 10–15 minutes, works with no account at all, and leaves your family with one habit worth keeping: sounding sure is not the same as being right.

Time and age: 10–15 minutes together tonight (add an optional deeper book check if you want to stretch it to 20), ages 6–12, with a Penguins card for ages 6–8 and a Chameleons card for ages 9–12.

Bring: a sheet of paper, a red and a green pencil, and one trusted reference you already own (an animal encyclopedia, a science picture book, or a library/museum website you trust). That’s it. The ready-to-read card above means you don’t need ChatGPT or any other AI account.

Before you start: two minutes for the parent

  • No account needed. The 3-sentence cards above are written by us for this activity; they are teaching examples, not live AI output. They work exactly like a real AI answer for practice, and you can finish the whole activity on paper.
  • If you want a live AI round, an adult runs it. OpenAI’s terms require ChatGPT users to be at least 13 (ages 13–17 need a parent or guardian’s permission). For ages 6–12, use your own adult account on your own screen: you type, your child watches, speaks the topic, and marks the paper. Never create an account for a child under 13 or enter a false birthdate.
  • What never goes into the chat box. Only general nature, science or history topics. Never your child’s name, school, city, photos or voice.
  • When to skip it tonight. If your child is tired or rushing homework, save it. The fun is in pausing to guess before anyone checks.

1. 📝 Predict first: what do we already know? · about 3 minutes

Parent, read aloud: “Tonight you’re the teacher and the AI is the student. AI has read a huge amount of text, so it always sounds confident, even when it mixes things up. Before we read its homework, tell me one thing you already know about penguins (or chameleons).”

Child: Say or write one fact you already believe is true. Ages 9–12: also name one detail you think someone could easily get wrong.

In your hands: One prediction on paper, made before any screen or card. Your child starts as the judge, not the audience.

2. 🔍 Read the three sentences and circle the sneaky one · about 5 minutes

Parent, read aloud: “Here’s the student’s homework: three sentences. Two are true and one is a mistake that sounds believable. Which one makes your teacher radar beep?”

Read the card above for your child’s age out loud. Keep the Parent Answer Key closed until your child has decided.

Child: Mark ✓ next to sentences you trust and ? next to the suspicious one. Say why before anyone checks: “I think sentence 2 is wrong because…”

Optional live round (adult-operated): copy the adult ChatGPT prompt from the card, paste it into your account, and read the three sentences aloud. Real AI doesn’t always follow “include exactly one mistake”. If it slips in two mistakes or none, that’s a great lesson too.

In your hands: A marked-up sentence list and a child who committed to an answer with a reason.

3. 📚 Check it in a real source and write the correction · about 5 minutes

Parent, read aloud: “Good teachers don’t ask the student ‘Are you sure?’, because the student will just sound sure again. Let’s check a real book.”

Child: Find the detail in your reference, then open the Parent Answer Key together and compare.

  • Ages 6–8: Say or write one fix: “Good try, AI! Actually, …”
  • Ages 9–12: Write the corrected fact plus the source name, and one line on why the mistake sounded convincing (for example, mixing up two cold places, or two kinds of lizards).

In your hands: One caught mistake, one correction backed by a real source, and a family phrase for next time: “Sounds sure? Let’s check.” Tap We tried it below. You don’t need to tick every box first.

Optional: a deeper book check (up to 20 minutes total)

Pick a second topic your child loves, ask the adult-operated AI for three new sentences, and repeat Steps 2–3. Ages 9–12 can add a challenge: paste the correction back to the AI and check whether its apology slips in a new exaggeration.

When it gets tricky

  • All three sentences look true: Break each one into who / where / how many and check the most specific place or number first.
  • The live AI made two mistakes, or none: Celebrate it. Your child just learned that AI doesn’t even follow “make one mistake” instructions perfectly. Switch back to the ready-made card for a clean round.
  • Your child wants to keep chatting with the AI: Close the screen and go back to paper. “The screen guesses words; our book and our evidence decide the grade.”

Three questions to end on

  1. What did AI do? Did it know the fact, or put familiar words together into a confident sentence?
  2. What did you do? Which clue made you suspicious first?
  3. How did we check? Which book or source settled it, and what will we do the next time an AI answer sounds 100% sure?

Use Print task checklist below for a paper copy. The completion certificate unlocks after you tap We tried it.

Lista de tarefas

O progresso fica neste navegador e dispositivo. O projeto gratuito não precisa de conta nem de envio de trabalhos.

1 tap to mark complete — no need to tick all 3 boxes first

Didn't finish tonight? What got in the way?

Optional 1-tap note · skip anytime
Where This Leads · Free Next Step & Parent Toolkit

Keep Going After Tonight: Free 3D Build & Parent Playbooks

Tonight’s activity is 100% free. Here is your next free hands-on build, plus real previews of the parent playbooks and printable cards in our annual library:

Free Next Step · 3D Dream House (~30 min, ages 5–12)

🏠 A casa dos nossos sonhos

Uma casa de blocos 3D, uma experiência com Magic e uma construção guardada localmente.

Continue with Free 3D Dream House →

What 10-minute family AI lab should we build next?

1-tap parent vote · no email needed

Lista de tarefas

A IA está certa? Seja o pequeno professor dela

O progresso fica neste navegador e dispositivo. O projeto gratuito não precisa de conta nem de envio de trabalhos.

  1. □ 📝 Say or write one thing we already know about the topic before reading the card
  2. □ 🔍 Read the three sentences together and circle the one your child thinks is wrong
  3. □ 📚 Check one trusted book or site together and say or write a one-sentence correction

Três perguntas sobre IA

O que fez a IA?

O que fizemos nós?

Onde é que a IA se enganou?