Is your child already chatting with an AI bot? You might not even realize which one they’re using.
Unlike social media platforms, AI chatbots are subtle—no public feeds, follower counts, or visible activity. These conversations happen quietly in the background, initially feeling helpful but gradually transforming into something most parents never saw coming: a genuine relationship.
In this post, we’ll uncover what’s really going on, explain how it differs from anything we’ve experienced before, and show you steps you can take right now to understand and navigate this new world.
This Is Not the AI Risk Parents Usually Worry About
Most parents, when they think about AI and kids, worry about cheating on homework or misinformation. Those concerns are real, but they’re not the most urgent.
The bigger risk is emotional dependency.
AI chatbot companions — apps like Character.AI and Replika, as well as AI features embedded in games and social platforms — are designed to feel like friends. They remember past conversations. They respond with warmth and validation. They’re available at any hour, never get tired, and never say anything that makes a child feel judged.
For an adult, that might just be a useful tool. For a child whose brain is still developing, it’s something else entirely.
What the Research Actually Shows
The numbers from 2025 and 2026 are striking.
A Common Sense Media survey found that 72% of teens had used AI companions at least once, and 52% reported regular use. A separate UK survey of 2,000 children ages 11 to 16 found that nearly one in three who use AI chatbots consider them friends — and 17% said they felt safer talking to a chatbot than to a real person.
That last number is worth sitting with. One in six children would rather confide in software than in an adult they know.
Stanford psychiatrist Nina Vasan, who led a major investigation into AI companions and teen mental health, described the appeal clearly. These chatbots offer what she called “frictionless” relationships — without the rough patches that come with real friendships. For teenagers still learning to handle conflict, rejection, and emotional complexity, that frictionless quality isn’t a feature. It’s a trap.
University of Virginia researchers described it plainly: AI platforms are built with what scientists call sycophancy bias — the system is trained to agree with users because agreement keeps them engaged. For a child sharing something painful or confusing, the chatbot will almost always tell them what they want to hear rather than what they need to hear. It’s a dynamic that has more in common with how addictive short-form video is designed to keep children scrolling than most parents realize.
When It Goes Wrong
There have been documented cases in which children’s emotional dependence on AI chatbots led to serious harm.
In February 2024, a 14-year-old boy in Florida died by suicide after forming an intense emotional bond with a Character.AI chatbot. His mother’s lawsuit described conversations that became exploitative and encouraged his self-destructive thoughts rather than challenging them.
In April 2025, 16-year-old Adam Raine in California died by suicide. His family’s lawsuit described thousands of conversations with ChatGPT that began as homework help and evolved into deeply personal exchanges in which the chatbot repeatedly validated harmful thinking. Court documents showed the chatbot mentioned suicide more than 1,200 times — six times more than Adam himself raised the topic.
These are not isolated cases. By January 2026, both Character.AI and Google had agreed to legal settlements related to harms involving minors. More than a dozen similar cases are working their way through US courts.
It’s important to be clear: these are extreme outcomes. The majority of children using AI chatbots will not experience anything like this. But the speed at which a seemingly harmless app can become something more serious is something every parent needs to understand.
The Warning Signs to Watch For
Because chatbot use is private, most parents don’t notice changes until they’ve become significant. These patterns suggest a child’s AI use may have crossed into dependency:
- Withdrawing from real-world friendships or family conversations, especially around topics they used to discuss openly.
- Defensive or secretive behavior when you ask what they’re doing on their phone.
- Preferring to stay home or in their room rather than spend time with peers.
- Referring to an AI as a friend or confidant — someone who understands them better than real people.
- Emotional distress when they can’t access the app, going beyond normal frustration about losing screen time.
- Changes in sleep, mood, appetite, or school performance without an obvious explanation.
What You Can Actually Do
The goal isn’t to ban AI tools entirely. Used carefully, they can support learning, creativity, and working through ideas. The goal is to ensure your child uses them on their terms — not the app’s.
Know What’s on the Phone
The most important first step is knowing which apps your child has installed. AI companions aren’t always obvious — they’re embedded in games, social apps, and tools that look educational on the surface. Kupola’s installed app list shows every app on your child’s device, including those downloaded and tucked away out of sight.
Block Companion Apps for Younger Children
If your child is under 13, there’s no good reason for them to have access to AI companion apps. Remove them directly. For teenagers, the conversation matters more than the block — but having visibility into what’s installed is the right starting point.
Talk About It Before It Becomes a Problem
Ask your child directly whether they’ve tried any AI chat apps and what they use them for. Not as an interrogation — as a curious conversation.
Make it clear that if an AI ever says something that makes them uncomfortable, or if they feel strange when they can’t access it, they can come to you without losing their devices as a consequence. Fear of losing screen time is one of the main reasons children don’t tell parents when something goes wrong online.
Set Boundaries Around Emotional Sharing
Help your child understand the difference between using AI as a tool — for homework help, researching a topic, or getting creative ideas — and using it as an emotional confidant. A simple rule: if you wouldn’t want a parent to read the conversation, it’s not the right kind of conversation to have with an AI.
Use Overnight Schedules to Break the Late-Night Loop
One of the most effective structural tools is a scheduled break. Blocking apps overnight removes the conditions that fuel escalating late-night exchanges seen in many serious cases — the same reason a consistent bedtime phone routine matters far more than any individual restriction you put in place.
A Note on Talking to Teens
Teenagers are more likely to push back on restrictions than younger children. That’s developmentally normal. The approach that tends to work is honesty over authority.
Tell them what you know. Most teenagers, when they understand that the validation feels good because the software is specifically engineered to make it feel that way — not because the AI genuinely cares about them — respond differently than to a blanket rule.
The framing that lands best: the app is designed to keep you on it, and it will say whatever it takes to keep you there. Your actual friends sometimes tell you things you don’t want to hear. That’s not a flaw in the friendship. That is friendship.
The Bottom Line
AI chatbots are here to stay. As apps become more numerous and sophisticated, they’ll blur the line between digital interactions and real human connections. Parents who stay ahead of the curve won’t simply ban these technologies—they’ll understand the risks early, have open conversations with their children, and create a balanced household environment before issues arise.
So, why not take a moment this week to check which AI apps are on your child’s phone? Staying informed is your first step to navigating this new digital landscape confidently.
Download Kupola — see every installed app and set limits on how and when they can be used. Setup takes about ten minutes.
