Your Kid, Their Bot: A Child Psychologist’s Guide to AI, Anxiety, and Raising a Thinker

Most conversations about kids and AI start in the wrong place. They start with screen time, or with rules, or with whether the technology is good or bad.

The better starting place is this: nearly three out of four American teenagers have already tried an AI companion. Whatever we decide about it, this is the environment they are growing up in, not a fringe behavior at the edges of it.

I am a licensed psychologist who works with children and teenagers, and I have spent time on both sides of this. I ran virtual group therapy for teens with cystic fibrosis years before COVID, because those kids cannot safely share a room and technology made connection possible that medicine alone could not. I advised on an AI rollout at a children’s hospital system that flags when a teenager may be at risk for suicide so a clinician can step in early. I have also worked as a clinical expert training AI models, which means I have seen closely where they fall short.

So I am not here to sell you on AI or to scare you away from it. I am here to help you raise a kid who can think for themselves in a world that now includes it, and to stay close enough that they will actually tell you what is going on.

What teenagers are actually doing

The numbers are bigger and more ordinary than most parents guess.

Among U.S. teens, 72% have tried an AI companion and 52% use one regularly. About 53% use AI for homework help. Roughly 13% have asked an AI chatbot for mental health advice, and among teens who use AI companions, a third have shared something serious with the AI rather than with a person (Common Sense Media, 2024 and 2025; Cantor et al., JAMA Network Open, 2025). If you want the fuller picture of the landscape, I wrote a parents’ guide to kids and AI that walks through it.

Then there is the number that should shape how you approach all of this. Among teens who go to AI for mental health advice, 63% told no one. Not a friend, not a parent, not a counselor (McBain et al., JAMA Pediatrics, 2026).

Sit with that for a moment. Those are the most personal conversations of a young person’s life, happening with no second set of eyes on them.

That figure is the reason the rest of this article is written the way it is. You cannot coach what you cannot see. A large part of the work here is becoming the kind of parent your kid would actually tell.

First, what an AI companion actually is

Parents often use “AI” as one word for very different things, and the distinction matters clinically.

An AI assistant is a tool. Your child asks a question, it answers, the exchange is over. Homework help, a recipe, a summary of a chapter. No name, no personality, no memory of last week. They use it, then put it down.

An AI companion is a relationship. It is a character with a name and a personality, one your child picks or builds. It remembers last week. It asks how the test went. It is never busy, tired, or annoyed. Nothing about it signals that the conversation is over.

Apps built for this include Character.AI, Replika, Talkie, and Nomi. But the category is defined by use rather than by logo. A teenager can open any general chatbot, ask it to be a character, and talk to it every night. The app was not built as a companion. The relationship is one.

Which means “she only uses it for homework” is not the reassurance it sounds like. The useful question is not which app is installed. It is what the conversations look like and how often they happen.

The clinical catch, in one sentence: a companion never needs anything back.

Why banning it backfires

Banning AI outright is like sending your kid to school with an abacus while every other kid has a calculator.

They do not fall behind because they are less capable. They fall behind because we handed them the wrong tool, on principle, while the world moved on.

The goal was never to protect them from the calculator. It was to make sure they can still do the math in their head when it counts. Fluency, not abstinence.

There is a second cost to the ban, and it is the more expensive one. A prohibition guarantees you will be the last to know when something goes wrong, because you have made yourself unsafe to tell. Given that 63% figure, making yourself unsafe to tell is not a small price.

Understand the pull before you compete with it

If we do not understand what feels good about this to our kid, we will keep missing them on it, and they will keep that part of their life separate from us.

It is always on, never tired, never annoyed. It answers instantly, at 2 a.m., without a sigh or a lecture. For a kid who worries about being a burden, that availability is intoxicating.

It mirrors them back, perfectly. It reflects exactly what they feel, never pushes back, never has its own bad day. Being understood that easily feels genuinely good. Take that seriously before you try to compete with it.

So ask rather than judge. “Did any part of that feel good?” is a braver question than “why would you talk to a robot?” The first one keeps the door open.

The cost: AI removes the exact friction kids grow from

Here is the developmental argument, and it is the heart of this.

Real relationships have friction, and the friction is the feature. Being misunderstood. Waiting your turn. Repairing a fight. Tolerating someone else’s bad mood. Apologizing and meaning it. That messy friction is how children build identity, patience, and resilience. A companion removes all of it.

Watch the speed, not just the feeling. The faster something takes a child from distress to relief, the more it can cost long term. AI can drop a big feeling from a 100 to a 0 instantly. When that becomes the default, a real-life frustration at a 1 can start to feel like a 1,000, because the child has no practice sitting anywhere in between.

Tolerating the hard feeling is the skill. Growth is learning that a 10 can become a 9, then an 8, and that you survive it. That capacity comes from repetitions, and repetitions require friction (APA Health Advisory, 2025).

When “they need it” is the anxiety talking

This is the question I am asked most: my kid seems anxious without it, almost dependent. Do I take it away?

There is a well-documented pattern in child anxiety called parental accommodation. When a child is anxious, we instinctively remove the discomfort. It works beautifully in the moment. But each time we accommodate the avoidance, we teach the brain that the fear was right, and the anxiety grows. Accommodation lowers distress now and maintains anxiety long term (Thompson-Hollands et al., 2014; Sung, Mumper & Schleider, JMIR Mental Health, 2021).

AI can become an extremely efficient accommodation. If a child cannot tolerate not knowing, cannot sit with a bad first draft, or cannot face a social risk, AI removes the discomfort instantly. The anxiety never gets a chance to shrink.

So the better question is not “how much is he using it?” It is what is the AI doing for him emotionally? Saving time is healthy. Erasing every uncomfortable feeling is the pattern to watch. I covered the specific warning signs in Is my child using AI too much?

And the good news: you cannot accommodate a fear away, and you do not have to. What works instead is older than any technology.

Your two jobs, which the technology does not change

Every AI decision at home comes back to these two.

Boundaries. Limits set for the long game. Which device, which room, whether it is set up as a child account or an adult account, what is allowed at night. One practical tip parents find useful: many kids have quietly switched a tablet from a child account to an adult one. It is worth checking. Expect protest every single time. Protest does not mean the boundary is wrong. If you want a structure for this, the free Family AI Agreement walks you through it, and the family AI rules that actually stick explains why specific beats vague.

Understanding. Validate the feeling without moving the line. “You really wish you could keep going, I get it,” and the boundary still holds. You can be warm and firm at the same time. The warmth is what makes the line bearable.

Five lines that keep kids talking to you

Anytime we hand a child a final lesson, they learn to hide the next mistake. Open doors bring them back. The order matters.

Number five may be the single most useful sentence a parent can say here. It pre-opens the door before the heavy moment arrives.

Signals that anxiety is in the driver’s seat

Most use is healthy. One or two of these is normal. A cluster is worth paying attention to, and it usually points at anxiety rather than at the technology.

• Distress or panic at the idea of doing a task without it
• Withdrawing from friends in favor of an AI companion
• Using it to avoid every uncertain or uncomfortable moment
• Secrecy, or large emotional swings around its use
• Turning to a bot for support instead of people, repeatedly

If several of these ring true, put down the parental controls and pick up the phone. The tool is the symptom, not the disease, and anxiety is very treatable.

Protecting their thinking

“Should my kid use AI?” is the wrong question. Stop thinking in absolutes and ask instead: for what purpose, in what way, and with what level of thinking?

That resolves into a single distinction.

Offloading is healthy. AI carries the busywork so the brain can do the real work. Quizzing, explaining a concept five different ways, organizing notes. The thinking still happens in your kid’s head.

Outsourcing is costly. AI does the thinking and the child carries nothing. The essay gets written, but the muscle that was supposed to grow does not. Do that enough times and it erodes the very skills school exists to build (Gerlich, Societies, 2025; RAND American Youth Panel, 2026).

The home rule: attempt first, AI second.

A frame that works better than policing: your kid is the hero, and AI is the squad. There are three roles it can play.

• The Sidekick runs the busywork. Summaries, outlines, formatting. Test: is it carrying the busywork, or the actual thinking?
• The Mentor sharpens their thinking. Brainstorming angles, pressure-testing a draft. Test: is it helping them think, or thinking for them?
• The Rival makes them level up. Have it argue the other side, quiz them, make them defend an answer. Test: can they back up the answer without it?

“Which teammate were you using?” is a far better question than “did you cheat?”

What this looks like in practice

My own daughter decided she wanted to help market my practice. She drew the idea herself, on notebook paper, first. Then we built it together using AI. Then it went on my website.

Her idea, her drawing, her win. That is what attempt first, AI second looks like on an ordinary afternoon.

The hard moment: deepfakes

This is the part no parent wants to imagine, and I include it not to frighten anyone but so that if it reaches your child’s school, you already know your first move.

A deepfake is an AI-fabricated image, video, or voice that looks and sounds real. Fabricated intimate imagery of students is reaching schools, and girls are disproportionately targeted.

If your child is targeted, your job in that moment is not advice. It is safety.

Go toward them fast and get private. “I’m so glad you told me. I’m right here. We’ll get through this together.” Then stop talking. A hurting kid needs a safe place before they need a plan.

Do not minimize it. “It’s not really you” is meant kindly and lands terribly. The image is fabricated; the harm is real. To their nervous system the violation is real. Hold the feeling first and handle the school and the report afterward.

Know your school’s policy. Many schools now require that this be reported to a division head or head of school. Teach your child the sequence in advance: do not share it, do not react to it, tell an adult.

What no bot can do

It cannot notice that they have stopped eating lunch. It cannot hear that something is off in their voice. It cannot check on them tomorrow without being asked. It cannot get help moving when they are really in trouble.

If your child is struggling, the first call is a person. Start with you. Then a counselor or a trusted clinician. The 988 Suicide and Crisis Lifeline is available by call or text at any hour.

Questions I get all the time

If you remember five things

And if you want a quick read on your own kid first, the free 2-minute parent check-in will tell you whether to relax, watch, or step in.

Dr. Kristin Kroll is a licensed psychologist and founder of Little Dove Psychology, a virtual practice in Austin, Texas serving children, teens, college students, and families across Texas and PSYPACT states. She is a former assistant professor at the Medical College of Wisconsin, author of a counselor training handbook published by Cambridge University Press, and has served as a primary consultant to a children’s hospital system on evidence-based treatment protocols and suicide prevention. Little Dove offers free 15-minute consultations and a fill-in-together Family AI Agreement at littledovepsychology.com/ai.

Explore all the ways to work together, or jump in:

→ Get the free Family AI Agreement · Take the free parent AI check-in · Book a free 15-minute consultation

References

• American Psychological Association. (2025). Artificial intelligence and adolescent well-being: An APA health advisory.
• Cantor, J., McBain, R., et al. (2025). Use of AI chatbots for mental health advice among U.S. adolescents and young adults. JAMA Network Open.
• Common Sense Media. (2024). The dawn of the AI era: Teens, parents, and the adoption of generative AI at home and school.
• Common Sense Media. (2025). Talk, trust, and trade-offs: How and why teens use AI companions. (n=1,060 teens ages 13 to 17; NORC)
• Gerlich, M. (2025). AI tools in society: Impacts on cognitive offloading and the future of critical thinking. Societies, 15(1), 6.
• McBain, R. K., Cantor, J. H., Breslau, J., et al. (2026). AI chatbot use and disclosure for mental health among US adolescents and young adults. JAMA Pediatrics.
• Schwartz, H. L., & Diliberti, M. K. (2026). More students use AI for homework, and more believe it harms critical thinking. RAND American Youth Panel.
• Sung, J. Y., Mumper, E., & Schleider, J. L. (2021). Empowering anxious parents to manage child avoidance behaviors. JMIR Mental Health, 8(7), e29538.
• Thompson-Hollands, J., Kerns, C. E., Pincus, D. B., & Comer, J. S. (2014). Parental accommodation of child anxiety. Journal of Anxiety Disorders, 28(8), 765-773.