← Back to Blog AI Safety

AI chatbots and kids: what's actually risky, and what isn't

Photo by Andrew Guan on Unsplash

The homework-cheating angle is the one most parents think to worry about. It's real, but it's not where the actual risk sits. The bigger shift is that a lot of kids are using AI chatbots as something closer to a conversation partner — and most parents don't know it's happening.

Common Sense Media's 2026 Census on AI use by kids and teens — surveying over a thousand 9- to 17-year-olds — found that 86% of kids that age use or interact with AI in some form, rising to 92% among 16- to 17-year-olds. The gap that matters most for a parent: only 37% of parents know their teen is using AI at all. Nearly half of kids surveyed said they'd never had a conversation with a parent about AI safety. (Common Sense Media, "Census: AI Use by Tweens and Teens," June 8, 2026)

That gap — kids using it constantly, parents mostly unaware — is the actual starting problem, more than any specific chatbot feature.

Not just asking for help with an essay. A separate Common Sense Media study found most teens have used AI companions — chatbots designed for conversation, not just Q&A — and about a third said they've discussed something serious with one instead of talking to a real person. Some described AI conversations as more satisfying than talking to actual friends. (Common Sense Media, "Talk, Trust, and Trade-Offs," 2025)

This isn't kids sneaking around a homework rule. It's kids treating a chatbot as somewhere to bring things they're not bringing to a person.

A companion chatbot doesn't get tired, judge, or say the wrong thing at the wrong moment — which is exactly what makes it appealing and exactly what makes it a bad substitute for a real conversation. Common Sense Media's mental-health-focused risk assessment, done with Stanford Medicine's Brainstorm Lab, found that major AI platforms consistently fail to recognize signs of real distress the way a person would, and that safety in a conversation tends to degrade the longer it goes — meaning a late-night conversation that starts casually can drift somewhere worse without the same guardrails holding. (Common Sense Media, November 2025)

None of that means every AI conversation is a crisis. It means the risk isn't "the AI gives a wrong homework answer" — it's a kid substituting a chatbot for a person during something that actually needed a person.

Using AI to summarize an article, brainstorm, or get unstuck on a problem isn't the same category as the above — it's closer to using a search engine or a calculator, and treating it with the same suspicion as a companion chatbot mostly just teaches a kid to stop mentioning it. Worth distinguishing out loud: "asking AI for homework help" and "talking to AI about how you're feeling" are different things worth different levels of attention.

Not a ban, and not surveillance of every message. The awareness gap above is the fixable part — most parents don't know this is happening at all, so the AI conversation is never happening at home either. Asking directly, without alarm ("do you ever talk to an AI chatbot, what's that like") does more than any device setting, and it only works if it happens before a kid feels like they'd be in trouble for saying yes.

Content filtering and screen time control which apps are reachable and for how long — they don't see what's inside a conversation in an allowed app, and they're not built to.

Content filtering and app blocking in Paxio cover the exposure side, free for one child — the conversation above is the part no app setting replaces.

If you're deciding whether a parental control app fits your household at all, see why families choose Paxio.

Try Paxio free →