Nearly 1 in 5 young people turn to AI for mental health advice. What Utah parents should know

SALT LAKE CITY — Artificial intelligence is increasingly becoming more than a homework helper or search tool for young people. For some, it is also becoming a place to talk about their mental health.

Nearly 1 in 5 U.S. adolescents and young adults surveyed said they had used an AI chatbot for mental health advice, according to a study published this summer in JAMA Pediatrics.

Perhaps more striking: Most of those young people had not told anyone they were doing it.

The nationally representative survey of 1,009 people ages 12 to 21 was conducted in November 2025. Researchers estimated 19.2% had used AI chatbots for mental health advice, up nearly 50% from the previous year.

Among those who had turned to AI for mental health advice, 42.8% said they did so at least monthly and 91.7% described the advice they received as somewhat or very helpful.

But 63.3% said they had not disclosed their use of AI for mental health advice to anyone.

For Dr. Todd Thatcher, a psychiatrist and vice president of medical services for Valley Behavioral Health, those numbers raise an important question for parents: How do you talk to a child who may already be confiding in artificial intelligence?

Related:

When you're talking to a chatbot, who's listening?

As the tech sector races to develop and deploy a crop of powerful new AI chatbots, their widespread adoption has ignited a new set of data privacy concerns among some companies, regulators and industry watchers.

Don't start with a lecture

Thatcher said parents should resist the urge to immediately confront a teenager or turn the conversation into a warning about technology.

"The last thing you want to do is lecture," Thatcher said. "It's not a lecture. It's a time to connect. It's a time to learn."

Instead, he recommended asking questions and listening.

One way to start is simply mentioning the research at the dinner table and asking children what they are seeing among their peers.

"What are your friends doing? What are you seeing at junior high and high school?" he suggested.

That approach can make the discussion less confrontational than directly asking a child whether he or she is using a chatbot for mental health advice.

Parents also shouldn't expect to settle the issue in one conversation. Getting a teenager to say a few sentences about what friends are doing may be enough for the first discussion. Parents can then find another opportunity to revisit the subject later.

He also recommends asking young people what they think AI does well and where they believe it could cause problems.

Is talking to AI about mental health always bad?

Thatcher cautioned against treating all use of AI for mental health as equally dangerous.

He said existing research suggests chatbots may have some usefulness for adults experiencing relatively mild mental health concerns, but their limitations become more significant as a person's needs become more complex.

He said specialized treatment for conditions such as obsessive-compulsive disorder and post-traumatic stress disorder should remain in the hands of trained professionals.

Thatcher is even more cautious when it comes to adolescents.

Young people's brains are still developing, including areas involved in judgment, emotional regulation and decision-making. A trained therapist can consider a teenager's development, circumstances and other factors while talking with them in a way a chatbot may not.

Thatcher said there is also another fundamental need at stake for adolescents: human connection.

"Teenagers are particularly sensitive to that and really crave and need that," he said.

When AI becomes something more

Thatcher said he recently experienced for himself how easily an interaction with AI can begin to feel personal.

Out of curiosity, he asked a chatbot to write a roughly 400-word obituary about him based on what it knew from their previous interactions.

He happened to be having a difficult day emotionally.

The response compiled information the chatbot had learned about him and presented it in a flattering way. Thatcher said reading it nearly brought him to tears.

"I really am a good person," he remembered thinking.

Then something happened the following day that caught his attention.

When he began feeling down again, Thatcher said a thought entered his mind: Go ask the chatbot how wonderful you are.

He stopped himself.

"That's the slippery slope down to an addiction to that and an inappropriate dependence on that," Thatcher said.

For him, the experience illustrated why adults shouldn't assume young people are the only ones susceptible to developing an emotional attachment to an AI system.

"We all have to deal with this," Thatcher said.

Thatcher said a chatbot can be appealing precisely because it is easy to access privately. But he encourages teenagers to recognize that convenience is not the same as connection.

"You need to make a human connection," he said. "It's far better than making a computer connection."

The stakes become higher in a crisis

The distinction becomes particularly important when a person is considering self-harm or suicide, Thatcher said.

He does not recommend relying on AI in those situations.

A 2025 study published in Scientific Reports illustrates some of those concerns. Researchers tested 29 AI-powered chatbot agents using standardized prompts designed to simulate increasing levels of suicide risk.

None met all of the researchers' initial criteria for an adequate crisis response. Fifteen met less-stringent criteria for what researchers classified as a "marginal" response, while 14 were considered inadequate.

Researchers identified problems including failures to provide appropriate emergency contact information and a lack of contextual understanding.

The study did find that most of the tested chatbots eventually recommended professional assistance or a crisis line. Researchers concluded the results raised concerns about using AI-powered chatbots in sensitive mental health situations without clinical validation.

Thatcher said thoughts of suicide, self-harm or harming another person should be treated as beyond the role of a chatbot and require human help.

What humans still offer

For Thatcher, the biggest difference between talking with a chatbot and another person can be difficult to quantify.

It's the experience of actually being with another human being.

"When you sit in front of another human being, they're going to make eye contact with you," Thatcher said.

A person can react emotionally, laugh with someone, recognize changes in behavior and ask questions based on what is happening in the moment.

"That's something that AI is never going to be able to reproduce," Thatcher said.

A therapist also brings lived and professional experience into the conversation and can help a person understand that what he or she is experiencing may not be as isolating or unusual as it feels.

Thatcher said that message can land differently when it comes from another person.

As AI becomes more commonplace in young people's lives, the authors of the JAMA Pediatrics study similarly concluded that parents and clinicians should proactively discuss chatbot use rather than assume it isn't happening.

For parents, it can begin without taking away a device, demanding to see a conversation, or trying to solve everything at once.

It can begin by asking a question and listening to the answer.

Related topics