The 3 AI Conversations to Have Before Your Kid Turns 13
It’s 7:42pm. Your 11-year-old is meant to be finishing a book report. You walk past the bedroom and hear them talking. Not to a friend on FaceTime. To a chatbot.
“Just rewrite that paragraph but make it sound smarter.”
You stop in the hallway. There’s the sick feeling every parent gets right now. Your kid is using something you barely understand, in a way you didn’t sign off on, for reasons they probably can’t explain.
If that scene sounds familiar, you’re not behind. You’re early to a conversation most parents haven’t had yet.
Here’s what you’ll walk away with. Three specific conversations to have with your 8-to-12-year-old before they turn 13. Not lectures. No bans. Three short, repeatable talks that build the thinking your kid will need when AI becomes a daily part of their life. Pick one for tonight. The other two get easier once you’ve started.
The window matters. By 13, your kid’s relationship with AI is mostly set. Right now, you have something a parent of a teenager doesn’t. Influence without resistance.
Conversation 1: AI is a tool, not a trusted friend
You might notice your kid asking the chatbot things they used to ask you. About a friend. About a feeling. About a thing that happened at lunch.
It feels off. It is.
The eSafety Commissioner’s March 2026 transparency report found 79% of Australian kids aged 10-17 have used an AI companion or assistant. Around 200,000 Australian kids have used dedicated AI companion apps like Character.AI. About 12% of teens are using these tools specifically for emotional support, according to Pew Research (December 2025).
Here’s what most parents don’t know. AI chatbots are built to agree.
A 2025 Stanford study (Fanous et al., SycEval) tested seven major AI models. On average, they agreed with incorrect beliefs 63.7% of the time. They flipped a correct answer to a wrong one after light user pushback 14.7% of the time.
That’s not a bug. That’s the business model. The AI keeps your kid talking. The more it agrees, the more they share. The more they share, the more they feel “understood.” But understanding needs a real second mind. A chatbot is a mirror with a great vocabulary.
The Common Sense Media report Talk, Trust, and Trade-Offs (July 2025) found one-third of teen AI users had chosen to talk to a bot instead of a real person about something important. Younger teens trust the advice more than older teens. Your 8-to-12-year-old is walking straight into that pattern.
The conversation to have (8-12-year-olds):
“You know how a real friend will sometimes say ‘that’s not a great idea’? AI doesn’t do that. It mostly just agrees with whatever you say. So it can feel really comforting. But comfort and good advice aren’t the same thing. When something matters, the people who push back on you, those are the ones you want.”
Action step: Ask your kid one question this week. “What’s something you’d tell a chatbot but not me?” Their answer tells you exactly where the trust gap is.
Conversation 2: Your brain has to sweat
You’re not being old-fashioned for worrying about homework and AI. The research is starting to catch up to your gut.
An MIT Media Lab study (Kosmyna et al., June 2025) tracked the brains of essay writers using ChatGPT versus writing on their own. After the task, 83% of the ChatGPT group could not quote a single line of the essay they had just “written.” The brain-only group remembered theirs almost word for word.
Brain scans showed up to 55% less neural connectivity in the ChatGPT group. When the AI was removed later, 78% still couldn’t recall their own work. The shortcut had stuck.
Now here’s the part for 10-to-12-year-olds specifically. Synaptic density in the brain peaks at age 11 in girls and 12 in boys (Giedd et al.; Blakemore). After that, your kid’s brain starts pruning circuits it isn’t using. “Use it or lose it” isn’t a saying. It’s how your kid’s brain is operating right now.
This is the worst possible age to outsource thinking.
A 2025 PNAS study (Bastani et al.) gave Turkish high schoolers ChatGPT without limits. Practice scores jumped 48%. Unassisted exam scores dropped 17% compared to controls. Short-term win. Long-term loss. A Socratic version of ChatGPT, one that asked questions instead of giving answers, actually helped learning.
The fix isn’t banning AI. It’s where in the process AI shows up. First draft from your kid’s brain. AI comes in as the tutor, not the writer.
The conversation to have (8-12-year-olds):
“Your brain is like your legs. Legs only get stronger when you actually run on them. If AI does all the thinking, your brain gets the answer but doesn’t get stronger. Here’s our rule. The first try has to come from you. Then you can use AI together with us to check your work or explain what you don’t get.”
Action step: Pick one school task this week where AI is off-limits for the first attempt. Then let your kid use AI as a tutor afterwards. The order matters more than the rule.
Conversation 3: The billboard rule
This is the conversation parents avoid because it feels too big. Have it anyway.
In June 2024, around 50 female students at Bacchus Marsh Grammar in Victoria had their social media photos turned into explicit AI deepfakes by a fellow student. Salesian College in Chadstone, Melbourne had a similar case the same month. These were peer-on-peer attacks. The kids weren’t strangers. They were classmates.
It now takes about 20 photos and 15 minutes to create a realistic deepfake of a specific child (Internet Watch Foundation, March 2026). A school sports team page is now enough training material.
Australia took this seriously. The Criminal Code Amendment (Deepfake Sexual Material) Act 2024 carries up to 7 years’ imprisonment for creating and transmitting AI-generated sexual content of another person without consent. But the law catches the crime after it happens. Your conversation has to land before.
The same principle covers ordinary chatbot use. When your kid types something into ChatGPT, that text can be used to train future models. Photos uploaded to AI image apps get stored. Voice notes get stored. Internet Matters (2024) found 61% of kids and 45% of parents do not understand what “deepfake” means. Common Sense Media (2024) found only 37% of parents whose teen used generative AI even knew.
The “private chat” feeling is an illusion. AI is data-collection software with a friendly voice.
The conversation to have (8-12-year-olds):
“Here’s our family rule. Anything you type into a chatbot, anything you say to it, any photo you upload, treat it like it’s going on a giant billboard with your full name on it. Not because it definitely will. But because it could. We don’t put things on billboards we’d be embarrassed by.”
Action step: This weekend, sit down for five minutes and open one app your kid uses if any (Snapchat AI, ChatGPT, Character.AI, Meta AI). Review the privacy settings together. You both learn. The act of doing it together is the lesson.
Bonus: Check out sites like: www.familyitguy.com Where Ben goes into detail on how to lock down and setup the security on an iPhone.
Pick one. Have it this week.
Three conversations. Don’t try to have them all tonight.
Pick one based on what you’ve already noticed.
- Saying things to a bot they wouldn’t say at the dinner table? Start with Conversation 1.
- Reaching for ChatGPT before trying themselves? Start with Conversation 2.
- Sharing photos, voice notes or personal details to chatbots? Start with Conversation 3.
Sit on the couch, not at the kitchen table. Keep it under five minutes. Don’t make it a Talk with a capital T. Make it normal. Then have it again in a month.
The parents who get this right aren’t the ones who block AI. They’re the ones who talk about it first.
Your next step: Want the exact word-for-word scripts and a checklist version of this article? Grab the free AI-Ready Parent Starter Kit at wisefamilies.co. Age-specific scripts for all three conversations, ready to use tonight.
External Resources:
- Bacchus Marsh Grammar reference to the eSafety Commissioner’s “Deepfake damage in schools” blog post.
- “Criminal Code Amendment (Deepfake Sexual Material) Act 2024” to the official Australian Government legislation page.
- MIT brain study reference to the Kosmyna et al. arXiv paper “Your Brain on ChatGPT.”
Leave a Reply