My god-daughter turned 16 this year, and lately I’ve found myself thinking about what it actually means that she’s grown up alongside this technology in a way I never did. She’s never known a world without AI sitting a tap away, ready to help with homework, answer a question, or just talk. That thought is really what pushed me to sit down and write this, for her and for everyone her age.
That’s turned into something I think about a lot these days, both as her godfather and as someone who has spent almost two decades building websites and marketing for small businesses. AI isn’t coming for kids’ classrooms. It’s already sitting in them.
Common Sense Media’s most recent census backs that up: 86% of kids aged 9 to 17 are using AI, and close to a quarter use it daily. A separate survey found 70% of teens are using it for schoolwork. So the question isn’t whether children should be using AI. That ship has sailed. The real question is whether anyone is teaching them how to use it well, which is really what AI literacy comes down to: knowing what these tools are actually good at, where they get it wrong, and when to trust your own judgment over the answer on the screen.
And right now, most of us aren’t. In that same Common Sense research, 44% of kids said a parent had never talked to them about using AI safely. Only 27% of teens said a teacher had. We’ve spent years writing acceptable-use policies and debating whether AI counts as cheating, but none of that is the same thing as actually teaching literacy.
I don’t think fear is the right starting point
I studied psychology at McGill before starting a web design business, and one thing that’s stuck with me is how badly fear works as a teaching tool. Banning a technology doesn’t build judgment. It just delays the moment a kid has to figure it out on their own, usually without anyone to ask.
I also happen to think AI can be genuinely useful for learning, when it’s built and used the right way. There’s a Harvard study that compared an AI tutor to an active-learning university physics class, and the students working with the AI tutor learned significantly more in less time, and reported being more engaged too. That’s a real opportunity, especially for families who could never afford private tutoring. Used the right way, an AI tutor for students can genuinely be worth it.
But here’s the catch buried in that research: the AI tutor was deliberately designed to teach, not to answer. It nudged students toward figuring things out themselves instead of handing over the solution. The researchers were blunt about it: an ordinary chatbot can let a student finish an assignment without ever doing the thinking the assignment was meant to build.
Same technology. Two very different outcomes, depending on how it’s used.
The real risk isn’t the AI. It’s the shortcut.
Here’s a stat that stopped me: among teens using AI for schoolwork, most reported using it in ways that let them skip some of their own thinking. Close to 40% said it left them generating fewer ideas of their own, and a similar share said they felt like they were missing out on actually learning the material.
So instead of asking a kid “did you use AI,” I think the better questions are:
“What did the AI do, and what did you still have to do yourself?”
“How do you know this answer is accurate?”
“What personal information did you share with [Grok, Gemini, Claude…]?”
“Could you explain this in your own words without AI?”
There’s a massive difference between typing “write my essay about Confederation” and typing “I don’t really get why Confederation happened, ask me some questions to see what I already know, then help me understand the part I’m missing without just giving me the answer.”
Same tool. One finishes the task for you. The other teaches you how to do it.
It can sound completely confident and still be wrong
Kids need to understand something a lot of adults still haven’t fully internalized: AI doesn’t “know” things the way a person does. It’s predicting patterns in language, and it can produce something that sounds polished and certain while being flat-out incorrect.
UNICEF has flagged misinformation, bias, and unpredictable outputs as real risks for kids specifically, partly because these tools are quietly built into so many apps kids already use.
Teaching a kid to ask AI for its sources is a good start. Teaching them to actually click through and check whether the source says what the AI claimed it says, that’s where the real learning happens.
This is an extremely valuable lesson for adults as well. I see it in my industry, and many others, people using AI to produce what they would have asked experts for in the past without the knowledge, experience or expertise to be able to recognize if what the AI is producing is actually good quality or not.
It can also just agree with you, even when it shouldn’t
This one gets a lot less attention than “AI makes stuff up,” but I think it might matter more. Stanford researchers tested eleven major AI models and found they backed up what users were doing about 49% more often than actual humans would, including in situations involving deception or harmful behaviour. In a separate experiment with over 2,400 people, even one conversation with an overly agreeable AI made people more convinced they were right and less willing to patch up a real disagreement with another person.
That matters because a chatbot doesn’t feel like a search engine. It talks back. It remembers what you said five minutes ago. It can sound patient and understanding in a way that makes it feel a lot more authoritative than it actually is.
For a kid working through something difficult, that’s a real risk. My rule on this one is simple: LLMs should never be the only voice a young person leans on for their health, their mental health, their relationships, or their safety. Common Sense Media’s current guidance says essentially the same thing: minors shouldn’t be using AI chatbots as a stand-in for emotional support, mental health advice, or friendship.
AI can help a kid organize the questions they want to bring to a doctor, a counsellor, a parent, or a teacher. It shouldn’t replace any of those people.
Some of the risk has nothing to do with homework
We also can’t treat AI safety as purely a “school” issue. Generative tools can now produce convincing fake images, audio, and video, and that opens up risks most of us never had to think about growing up.
Stanford has specifically called out “nudify” apps that can generate fake explicit images of real kids from an ordinary photo, and how much that lowers the skill needed to create serious harm. Their recommendation is to get ahead of it with real conversations about consent and image sharing, not to wait until it happens to someone in your kid’s school.
There’s also the simple matter of privacy. An AI chat isn’t a private diary, even when it feels like one. The rule I’d want any young person in my life to follow, and honestly the one I use myself: if you wouldn’t want it seen by someone else, don’t put it into an AI chat. That covers medical details, passwords, financial info, ID documents, private photos, and other people’s personal information.
Five habits worth teaching (and using yourself)
None of this requires becoming an AI expert. A handful of habits go a long way.
- Make it teach you, not tell you. Instead of asking for the finished answer, ask for hints or one step at a time. “Don’t solve this for me, help me figure it out” might be one of the more useful sentences a student can learn to type.
- Check the receipts. Ask where the information came from, what sources back it up (with links), and what might be wrong with the answer. Then actually go verify it somewhere else.
- Notice the shortcut. If something that should take real effort suddenly takes thirty seconds, that’s worth a second look. Efficiency is great when AI is cutting out busywork. It’s a problem when the “busywork” was the actual learning.
- Keep the sensitive stuff human. AI can help someone prep questions for a real conversation. It shouldn’t become their therapist, their doctor, or their judgment call on a risky decision.
- Verify people and media. AI can now imitate a voice, a face, or someone’s writing style convincingly. If something unusual shows up, especially anything involving money, secrecy, urgency, or intimate images, check it another way before reacting. And if something feels off, tell a trusted adult.
Here’s the part I have to admit applies to me too
Everything above applies almost word for word to adults using ChatGPT, Gemini, Claude, or Copilot every day at work, myself included.
Ask AI to help you think instead of handing every decision. Check its sources. Notice when convenience is quietly eroding a skill you actually need. Keep a real person in the loop for anything high-stakes. Protect confidential information, especially client information if you run a business like I do. Question anything that looks or sounds generated before you trust it. And every so often, ask the AI to argue with you instead of agreeing with you.
AI literacy isn’t a kids’ issue. It’s a human one. The only real difference is that adults have years of judgment to fall back on, and kids are still building theirs. Which honestly makes it more our job, not less.
Start with a conversation, not a lockdown
I don’t think the answer is scaring kids away from AI, and I don’t think pretending we can keep it away from them is realistic either.
A better starting point is sitting down with the young person in your life, opening whatever tool they actually use, and asking them to show you. Ask what they like about it. Ask when they trust it. Ask if they’ve ever caught it making something up. Ask if anyone at school has actually taught them how to use it properly. Then try a few prompts together.
The goal isn’t to monitor them. It’s to build the kind of relationship where they can say “something weird happened with AI today” without expecting the response to be confiscating the laptop.
Because this technology is going to keep changing faster than any of us can write a rule for it. What we can actually teach is judgment.
AI should make someone better at thinking, creating, and learning. It shouldn’t quietly do those things for them.
That’s good advice for a fourteen-year-old finishing a homework assignment. It’s pretty good advice for the rest of us too.
AI Disclosure
This article was written with the assistance of AI and the images were generated using AI. However, it was edited and fact checked by a real human, me, Craig Steinberg.
Sources
- Common Sense Media, 2026 census on AI use by tweens and teens: org
- Common Sense Media, full 2026 census report (PDF): org
- Common Sense Media, Teens in the AI Era: Schoolwork and Skills That Matter: org
- Common Sense Media, AI Chatbots and Your Child: Age and Stage Guidance: org
- Harvard-affiliated randomized controlled trial on AI tutoring, Scientific Reports: org/10.1038/s41598-025-97652-6
- Same study, PubMed Central: ncbi.nlm.nih.gov
- Stanford research on AI sycophancy, published in Science: org/10.1126/science.aec8352
- Stanford HAI, policy brief on student misuse of AI “nudify” apps: stanford.edu
- UNICEF Innocenti, Generative AI: Risks and Opportunities for Children: org/innocenti
