Ask NOVA: learn how chatbots work, then check the answers

Students will be able to

1. Explain in their own words what a chatbot does when it answers a question.
2. Name one reason an AI answer can be wrong even when it sounds confident.
3. Check a claim against a cited source and label it confirmed, partly confirmed or not found.
4. Describe one habit for using an AI tool responsibly.

Materials

• A device per student or pair with a web browser (no account or sign-in needed).
• NOVA, our free AI tutor: aiunderstanding.org/tutor (the lesson link carries a campaign tag: /tutor?utm_source=hourofai_lesson&utm_medium=lesson&utm_campaign=hourofai2026).
• The printed NOVA Investigation Log, one per student, or the on-screen version below.
• Optional: a projector to show one conversation to the whole class.

Open NOVA in a new tab to try the lesson flow before class.

The plan

1. 0–5 min

   Warm-up: what happens when you ask?

   Teacher: Ask: when you type a question into a chatbot, what do you think it does before it answers? Collect three or four guesses on the board without correcting them.

   Students: Write one sentence on the Investigation Log: what I think a chatbot does.

2. 5–10 min

   Meet NOVA

   Teacher: Open NOVA on the projector. Point out the three modes (Explain, Socratic, Practice), the level picker (Kid, Beginner, Comfortable, Expert) and the note that NOVA can make mistakes. Set the level to Beginner for the demo.

   Students: Open NOVA on your device. Choose Explain mode and the level that fits you.

3. 10–25 min

   Ask and record

   Teacher: Give the class the first question: "How does a chatbot like ChatGPT actually work?" Circulate. Remind students that under every answer NOVA lists the guides it used, and that those guides are the evidence for the next step.

   Students: Ask NOVA the question. In the log, write three claims from the answer in your own words, and copy the title of one guide NOVA cites. Ask one follow-up question of your own.

4. 25–40 min

   Investigate the evidence

   Teacher: Students open the cited guide. For each claim they decide: confirmed (the guide says it), partly (the guide says something close), or not found. Ask pairs to compare verdicts and discuss any disagreement.

   Students: Open the guide NOVA cited. Check your three claims and mark each one. Then tap Quiz me and answer one question on what you just read.

5. 40–47 min

   Discuss: trust, but check

   Teacher: Lead a short discussion: which claims were easy to confirm? Did anyone find a claim the guide did not support? Why does NOVA say it can make mistakes, and what does that mean for how we use any AI tool?

   Students: Share one confirmed claim and one thing you would double-check before repeating it.

6. 47–50 min

   Exit ticket

   Teacher: Collect the Investigation Logs or read the exit-ticket answers aloud.

   Students: Answer the two exit-ticket questions at the bottom of the log.

Exit ticket

1. In one sentence: what does a chatbot do when it answers you?
2. Name one thing you would always check before trusting an AI answer.

Teacher notes

• NOVA is free and needs no account, so students never create a login. Conversations stay in the student's own browser; nothing is stored in a learner database.
• NOVA only answers questions about artificial intelligence. If a student asks about something else, it answers briefly and steers back to AI.
• Students who prefer to talk can use Voice mode; it works in browsers with speech recognition such as Chrome, Edge and Safari.
• Multilingual classes: NOVA answers in all 40 of the site's languages. Students choose their language from the flag in the site menu.
• Socratic mode is a good extension for fast finishers: NOVA asks them questions instead of answering.
• This is an independent AI Understanding lesson, not an official CodeAI / Hour of AI activity.