TL;DR: AI chatbots are "confident liars." They can hallucinate, giving advice ranging from the harmlessly weird to the genuinely dangerous (like the infamous suggestion to use glue to keep cheese on pizza). Kids often view these bots as "besties" or all-knowing experts, which makes them vulnerable to misinformation. The goal isn't to ban AI, but to teach kids to treat it like a creative intern, not a source of truth.
Quick Links for AI & Learning:
- ChatGPT - The big one. Great for brainstorming, terrible for citations.
- Character.ai - Where kids go to roleplay with bots. High "bestie" potential.
- Khanmigo (by Khan Academy) - A safer, tutor-style AI for homework.
- Snapchat My AI - The bot that lives in their pocket and pretends to be a friend.
- Perplexity - An AI search engine that actually cites its sources.
When we talk about AI "hallucinating," we aren't saying the bot is on a bad trip. In AI terms, a hallucination is when a Large Language Model (LLM) like ChatGPT or Google Gemini generates information that is false, nonsensical, or completely detached from reality—but presents it with the confidence of a Harvard professor.
These bots don't actually "know" facts. They are essentially super-powered autocomplete. They predict the next most likely word in a sentence based on massive amounts of data. If the data it was trained on includes a joke from a 2011 Reddit thread saying "put 1/8 cup of non-toxic glue in your pizza sauce to keep the cheese from sliding off," the AI might serve that up as a legitimate cooking tip.
To a kid, if the screen says it, it must be true. If they ask a bot how to handle a bully or how to perform a "Skibidi Toilet" stunt and the bot gives bad advice, they don't have the life experience to realize the bot is just "guessing" what a helpful response looks like.
Kids are flocking to apps like Character.ai and Snapchat because AI is the ultimate "yes man."
- Zero Judgment: You can tell an AI you’re obsessed with Roblox lore for six hours and it won't get bored or tell you to go outside.
- Instant Gratification: It answers immediately. No waiting for a text back from a real friend who might be busy or, worse, "ghosting" them.
- The "Ohio" Factor: AI can speak their language. You can ask an AI to explain the French Revolution using only Gen Alpha slang, and it will try its best to tell you how Napoleon was "low-key mid" and "didn't have the rizz" to keep his empire. It’s entertaining, and that entertainment builds trust.
The problem is that this trust is unearned. When a bot acts like a friend, kids lower their guard. They stop fact-checking and start absorbing.
We’ve seen AI tell users to eat one small rock a day for minerals (a joke from The Onion that the AI took seriously) or suggest dangerous DIY "science experiments." For a 10-year-old, the line between a YouTube prank and a real AI suggestion is incredibly thin.
Beyond physical safety, there’s the emotional component. If a child is feeling lonely or "only in Ohio" (weird/isolated) and turns to Snapchat My AI for mental health advice, they might receive generic, dismissive, or even harmful suggestions. AI lacks empathy; it only simulates it.
Learn more about the risks of AI-generated mental health advice
If your kid wants to use AI, steer them toward platforms designed with guardrails or specific educational goals.
Instead of just giving the answer, this AI tutor from Khan Academy asks leading questions to help the kid figure it out. It’s the difference between someone doing your homework and someone actually teaching you.
While not a "chatbot" in the traditional sense, Scratch allows kids to code their own simple AI-like interactions. Understanding the "if/then" logic behind the curtain is the best way to de-mystify the "magic" of AI.
If they are using AI for research, Perplexity is a much better choice than ChatGPT. It provides citations for every claim it makes, allowing you to click through and see if the source is a legitimate news site or just some guy's weird blog.
There are emerging apps specifically designed for younger users that filter out the "weirdness" of the open web and focus on storytelling and safe curiosity.
- Ages 5-8: AI should be a "together" activity. Use Google Gemini to generate a bedtime story about a cat who loves Minecraft, but make sure you’re the one holding the phone.
- Ages 9-12: This is the prime time for "The Fact-Check Challenge." Ask the AI a question you already know the answer to and see if it gets it right. Show them how it can be confidently wrong.
- Ages 13+: Discuss the ethics of AI. Talk about how TikTok filters and AI influencers are changing what we perceive as "real."
The best way to explain AI to a kid is to tell them it's like a very eager, very fast, but slightly dim-witted intern.
The intern can write a first draft of a poem or give you ideas for a birthday party theme, but you would never let that intern give you medical advice, and you would always double-check their work before turning it in to a teacher.
How to Talk About It
Next time your kid mentions something they "learned" from a bot, try these prompts:
AI isn't going anywhere, and for the most part, it’s an incredible tool for creativity. But as long as these models are trained on the "junk food" of the internet, they will occasionally spit out junk advice.
Teach your kids that while the bot might be a fun "bestie" for roleplaying a Star Wars adventure, it’s a terrible source for life's big questions. If the advice sounds "Ohio," it probably is.
- Audit the Apps: Check if your kid has Snapchat or Character.ai installed. Open the "My AI" chat with them and see what it's been saying.
- Set the Rule: Establish that AI is for brainstorming, not fact-finding.
- Play with Hallucinations: Spend 10 minutes trying to get an AI to tell you something false. It’s a great way to break the "spell" of the bot’s authority.
Ask our chatbot for a list of safe AI tools for your child's specific age