The talk that isn’t happening

Your child may already have a homework partner you have never properly met.

It sits in another tab while maths is being explained, an essay is being untangled, or a science topic is being turned into flashcards. It is patient, instant and often plausible. Many parents know it is there. Far fewer know what their child asks it, what the child accepts, or when the child decides, “That does not sound right.”

In a recent survey of 1,030 Australian parents who regularly use AI and have at least one school-aged child using it at home, 75% said their children use tools such as ChatGPT, Gemini or Claude for schoolwork, study help or research. Yet only 42% said they review or discuss how their children use AI.

That second number matters because the first one is already a fact of family life.

Confidence is not the same as contact

Parents in that survey were not indifferent. Their leading worries were the privacy and security of family information (58%), whether AI is accurate (47%), and whether family members might rely on it instead of forming their own judgement (45%). They were also broadly confident: 92% believed their family understands or uses AI as well as, or better than, other families.

Those feelings can all be true at once. A parent can use AI at work, worry sensibly about privacy, and still have no window into how a child uses it at 9:40 on a school night.

Confidence often comes from familiarity. We see the app open, hear that it “helped with homework,” and may even have shown our child how to use it. But familiarity is not oversight, and oversight is not the same as reading every message. The missing piece is usually a recurring, low-stakes conversation in which AI use is treated as neither a triumph nor a transgression.

The aim is not to catch children out. It is to stay in the loop while they learn to think about what the tool is doing.

Why conversation travels further than monitoring

Controls have their place, especially for younger children. They can limit access, filter some content and create breathing room while children mature. But a control mostly answers one question: can my child reach this?

It does not answer the questions that shape judgement:

• What was my child trying to do?
• Why did this answer seem convincing?
• What did they change, check or ignore?
• What would they do when the AI is unavailable, unhelpful or wrong?

Those habits matter because AI will not stay inside one approved app on one supervised device. It is appearing in search, documents, learning platforms, games and creative tools. Children also move between personal, shared and friends’ devices. A determined child may find a workaround; even a compliant child will eventually meet a tool you have not reviewed.

Digital-parenting research calls the alternative “active mediation”: rather than relying only on restrictions or after-the-fact monitoring, adults discuss what children encounter, sometimes use technology alongside them, and help them interpret it. Rules still matter, but conversation gives them a better chance of becoming internal judgement rather than external pressure.

It also protects trust. If every mention of AI leads to a confiscated phone, lecture or forensic inspection, children learn not to mention it. You want earlier disclosure, not better hiding.

That is why the best family AI ritual is not a surprise audit. It is an appointment.

Try a 10-minute “show me what you asked it” exchange

Choose a predictable weekly moment: after dinner, during a walk, on a regular drive, or while clearing up. Ten minutes is enough. The shorter and more ordinary it feels, the more likely it is to survive busy weeks.

Bring one example each.

Your child chooses an AI interaction from the week: a helpful explanation, a strange answer, a prompt that worked, or a moment when they decided not to use it. You bring one of yours: perhaps an email AI helped draft, a recipe it suggested, a summary it produced, or an answer you did not trust.

Then trade four quick reflections:

1. What was I trying to do?
2. What did the AI make easier?
3. Where did I pause, push back or change my mind?
4. What would I do differently next time?

Notice what is not on that list: “Prove this answer is correct.” We have written elsewhere about helping kids question AI, use a thinking checklist, and explain an answer back in their own words. Those skills examine a particular output. This ritual is different: it makes AI use discussable over time, before a concern becomes a crisis.

Let your child teach you something. They may show you a better way to narrow a prompt, request examples at the right level, turn notes into practice questions, or spot when an answer sounds polished but says little. If they reveal a shortcut you dislike, start with curiosity: “What problem were you solving?” You can still set a boundary. But a child who expects to be heard is more likely to speak up when something feels confusing, creepy or too good to be true.

With younger children, keep it concrete and demonstrate together. With older teenagers, respect the difference between oversight and reading a diary. Ask for the shape of the interaction — the question, useful part and doubtful part — rather than demanding an entire history. Agree what would make you look more closely together: secrecy, distress, contact from strangers, personal information being shared, or schoolwork submitted without understanding.

Five starters that do not sound like an interrogation

The first question sets the temperature. These are designed to invite a story, not a confession.

1. “What did AI save you time on this week — and what did you do with the time?” This treats AI as part of real life and asks whether saved effort became learning, rest or something else.
2. “Did it say anything that made you think, ‘Hmm, I’m not sure about that’?” You are rewarding doubt, not demanding proof. If they have never doubted it, that is useful to know.
3. “Show me a prompt that worked better than your first attempt. What did you change?” Better prompting is clearer thinking: naming the level, length and kind of help wanted.
4. “Was there a moment you chose not to use it? Why?” This normalises restraint and reveals where your child already draws lines — around creative work, friendships, tests or feelings.
5. “Teach me one AI trick — or warn me about one trap.” Reversing roles changes the power balance. Children explain more honestly when they are the expert.

Pick one. Leave gaps. Avoid turning a good answer into a speech. If the conversation lasts four useful minutes and ends naturally, that is success.

What “good” looks like after a month

You probably will not emerge with a complete map of your child’s AI life. That is not the test.

Good looks like your child saying, unprompted, “It gave me an answer, but it sounded made up.” It looks like them asking whether a photo, document or personal detail should be pasted into a chatbot. It looks like debating whether using AI to brainstorm differs from using it to write the paragraph. It looks like you discovering that your own first prompt is not always clear.

Most of all, it looks like AI becoming neither a forbidden room nor an invisible tutor, but a tool that can be shown to someone else and talked about without drama.

Start this week. One question. Ten minutes at most. You do not need to know more about AI than your child does. You only need to make it normal to ask: “Show me what you asked it — and tell me what you made of the answer.”