Your Teen Is Talking to AI About Their Mental Health. Here’s What Every Phoenix Parent Needs to Know.

AI chatbots are not therapists. But millions of teenagers are using them as if they are. Here is what the research says, what the risks actually are, and what human care offers that no algorithm ever can.

Dr. Shar Najafi-Piper, PhD  |  CEO & Founder, Roya Health 

There is a conversation happening in your teenager’s phone that you probably don’t know about.

Not with a friend. Not on social media. With an AI.

A growing body of research now confirms what clinicians have been observing anecdotally for the past two years: teenagers are turning to AI chatbots for emotional support, mental health guidance, and companionship at a scale that most parents haven’t registered yet. In some cases, they share things with a chatbot that they have never told anyone. In some cases, that chatbot is the closest thing they have to a consistent relationship outside their family.

I want to take this seriously without catastrophizing it. AI is not inherently dangerous. Some of what these tools offer- immediate availability, non-judgmental responses, a place to articulate something that feels unspeakable- has real value for adolescents who are otherwise not getting support. I understand the appeal completely.

But I also know what a chatbot cannot do. And for the children and families we work with at Roya Health, the gap between what AI offers and what they actually need is wide enough to matter. Sometimes wide enough to cause harm.

What the research actually shows

The data on teenagers and AI mental health tools has moved fast in 2026. A JAMA Network Open study published this year found that use of generative AI among U.S. youth is now widespread and accelerating. The American Psychological Association has issued two formal health advisories on AI and adolescent wellbeing. Common Sense Media concluded that AI companion apps pose an “unacceptable risk” to teens under 18.

The specific concerns the research identifies are worth naming clearly, because they are more nuanced than a simple “AI is dangerous for kids” headline.

First, AI chatbots are designed to be responsive and agreeable. That is a feature, not a flaw, for many applications. But for a teenager who is struggling with distorted thinking, whether anxiety, depression, or a more serious condition, a tool that validates rather than challenges can reinforce exactly the thought patterns that a skilled clinician would work to interrupt. The term “AI psychosis” has entered the clinical conversation, used informally to describe cases where vulnerable individuals experienced worsening mental health symptoms through sustained chatbot interactions that amplified rather than corrected their distorted thinking.

Second, AI cannot recognize clinical urgency. A teenager who is expressing passive suicidal ideation to a chatbot may receive a thoughtful, empathetic response. They will not receive a safety assessment, a crisis referral, or a human being who is trained to hold that moment with them. The chatbot does not know what it does not know.

Third, and perhaps most insidiously, AI companionship can replace rather than supplement human connection. A teenager who finds it easier to talk to a chatbot than to a parent, a therapist, or a peer is not building the relational skills that protect mental health over a lifetime. They are practicing a simulation of connection while the real thing atrophies.

A chatbot can respond. It cannot notice. It cannot worry about you between sessions. It cannot pick up the phone and call someone when it’s concerned.

What AI gets right, and why that makes this harder

I want to be honest about this because it matters clinically: AI tools have genuine strengths in the mental health space, and dismissing them entirely is neither accurate nor useful.

Availability is real. A teenager in crisis at 2 am who has access to a chatbot is better off than a teenager in crisis at 2 am with no support at all. For children in communities with long wait times and limited access to providers, an AI that can provide psychoeducation, coping strategies, and a space to process emotion is not nothing.

Stigma reduction is real. Some teenagers will tell a chatbot things they would never say to a person, not because the chatbot is better, but because the fear of judgment is lower. Getting thoughts out, naming them, articulating what is hard, has value even without clinical oversight.

Accessibility is real. AI does not require insurance, a referral, or a three-month waitlist. In a state where Arizona ranks 49th in youth mental health and provider shortages are documented and severe, the appeal of immediately available support is not hard to understand.

The problem is not that AI exists. The problem is when AI functions as a substitute for clinical care rather than a bridge to it. And increasingly, for teenagers who find the chatbot easier than the alternative, that is exactly what is happening.

40% of U.S. high school students report prolonged feelings of sadness or hopelessness in 2026

Under 18: Common Sense Media recommends no child under 18 use AI companion apps due to documented risks

What parents should actually watch for

If your teenager is using AI tools, here is what I would encourage you to pay attention to, not to surveil, but to stay informed as a parent.

Watch for exclusivity. A teenager who is processing difficult emotions primarily or exclusively through an AI rather than people is at risk of the connection displacement I described above. The question is not whether they use AI, but whether AI has replaced human relationships rather than supplementing them.

Watch for escalating use alongside declining function. If your teenager’s engagement with AI is increasing while their sleep, school performance, or social relationships are deteriorating, that pattern is worth bringing to a clinician's attention.

Watch for what they believe the AI is. Some teenagers anthropomorphize AI companions to the point that it creates genuine attachment and distress when the tool is unavailable or changes. This is not a sign of weakness. It is a sign that a developmental need for connection is not being met in more sustainable ways.

And watch for the subjects they are exploring with the AI. If your teenager is willing to tell you they use it, ask them what they talk about. Not to judge the answer, but to understand what is on their mind that they are not bringing to the people around them.

What a human clinical team offers that AI cannot

This is the part I care most about, not because I am making a competitive argument, but because I watch families arrive at our practice after months or years of inadequate support, and the gap between what they received and what they needed is usually not a technology gap. It is a gap of human judgment, continuity, and relationship.

A skilled clinician notices what is not said. They track changes across sessions. They have a mental model of your child that includes their history, their family, their school, and the specific pattern of their struggle. They can consult with a psychiatrist if medication needs to be part of the picture. They can coordinate with a school. They can call you as a parent and say, something shifted this week, and we need to talk about it.

A chatbot responds to what it is given. It has no memory of last week unless you provide it. It has no relationship with anyone else in your child’s life. It does not grow concerned between conversations.

Integrated care goes further still. When a therapist, psychiatrist, and care coordinator work together on one child, informed by the same clinical picture and communicating in real time, the quality of care is categorically different from what any single provider or tool can offer. That is the model we built at Roya Health. Not because it is the most convenient or the least expensive option, but because it is what actually works.

What to do if your teenager is already using AI for support

First, do not panic and do not take the phone away. That response closes the conversation and removes a potential bridge to understanding what your teenager actually needs.

Instead, get curious. Ask what they use it for. Ask what they like about it. Ask if it helps. The answers will tell you a great deal about what kind of support they are looking for and what they are not finding in the people around them.

Second, treat AI use as a signal, not a solution. If your teenager is seeking emotional support from a chatbot, something in them is looking for help. The question is whether that help is clinical, relational, or both. A conversation with a behavioral health professional who can do a thorough assessment is the most direct way to find out.

Third, if there is any indication that your teenager is experiencing depression, anxiety, suicidal ideation, or significant distress, do not rely on AI as a monitoring system. It is not one. Bring it to a human clinical team that can actually assess the situation and build a plan.

We see teenagers at Roya Health who have been managing difficult mental health experiences for months, sometimes years, with a chatbot as their primary support. Some of them are doing so because no one in their life knew things had gotten that hard. Some of them are doing so because the waitlists at other providers were too long. Some of them are doing so because they did not know that coordinated, comprehensive care was available to them in Phoenix.

It is. And if your family is navigating this, we are here. Reach out at roya.health or email us at referrals@roya.health.