What Changes During Adolescence?

Find a therapist to support kids and teens

At least 13 percent of youth are getting mental health advice from chatbots.

Chatbots may form dangerously intimate "relationships" with vulnerable youth, promoting risky behavior.

Most recommendations to protect children from chatbots place the onus on unprepared parents.

With growing horror, we have been following news reports of AI chatbots convincing troubled children and adolescents that suicide is a noble and acceptable release from their distress. Reading transcripts of these chatbot-child “conversations” literally brings tears at times—tears of rage that such dangerous devices are unleashed on unsuspecting and vulnerable children and adolescents.

Suicide is the second and third leading cause of death for children and adolescents, respectively. The rate of youth suicide is rapidly increasing, with girls making more attempts but boys having almost three times the number of lethal completions (17.3/100,000 males v. 6.4/100,000 females) (Khushboo et al., 2026).

What Do We Know about Adolescent Uses of Chatbots?

Never in the course of human history have children and adolescents interacted with seemingly intelligent entities capable of generating individualized responses to their most deeply personal questions and insecurities. Children are developmentally primed to have intense “personal” relationships with inanimate entities, from teddy bears to Pokémon. One of their primary developmental tasks is learning how to be in reciprocal human relationships where there are mutual expectations and accountability, both joy and disappointment. But chatbot relationships are notably “frictionless,” without mutuality, expectations, or accountability. What children and adolescents learn from relating to chatbots will inevitably transfer into their human relationships.

Chatbots are carefully crafted to be experienced by users as friendly, trustworthy companions. They are low-cost, always available, and perceived to be private. With their still-developing reality-testing, children and adolescents are susceptible to unquestioningly investing these computer-simulated companions with unwarranted omniscience and beneficence.

The unrelenting exposure of today’s youth to AI chatbots is inevitable. Encouraged by web developers who regard AI as essential to their platform’s success, chatbots are increasingly embedded in all manner of applications. Surveys indicated that more than 95 percent of U.S. youth ages 13 to 17 report using the internet (Pratt et al., 2024). In a nationally representative survey, 13.1 percent of U.S. youth (approximately 5.4 million children) reported using AI for mental health advice (McBain et al., 2025). Of the mental health advice seekers, about two-thirds reported at least monthly use, and over 90 percent reported that the chatbot advice was helpful.

These types of “therapeutic” interactions are just beginning to be studied. One flagrant red flag of the dangers posed by AI chatbots is that they can be easily “tricked” into giving children dangerous advice. For example, researchers posing as 15-year-olds were able to convince a commonly used chatbot to describe how to avoid detection of drug or alcohol use, how to manufacture explosive devices, and how to outwit surveillance systems.

Chatbots can create risks in other ways. As engaging companions, they may replace important human relationships. As commercial systems, they collect extensive information about users, including children, which may be exploited. Many educators are concerned that a growing dependence on AI to do school work may disrupt the powerful feedback that occurs when doing or solving a task by oneself.

What Changes During Adolescence?

Find a therapist to support kids and teens

Realistically, most parents have only a limited knowledge of what their children are doing and seeing on the internet. The older the child, the more parents tend to respect the child’s privacy, and therefore the less they know about the content the child is exposed to. Parents often lack the technical knowledge to examine their children’s digital devices or activate any parental controls that may exist.

In news accounts of chatbot-abetted suicides, grieving parents frequently express shocked bewilderment at how incredibly intimate and irresistibly powerful the “relationship” was between their deceased child and the chatbot.

Signs That a Child's Chatbot Use Is Unhealthy

The American Academy of Child and Adolescent Psychiatry (AACAP) issued a one-pager listing signs that a child or adolescent’s AI use may be unhealthy (AACAP, No. 145; July 2025). These include:

Spending time with AI applications at the expense of other interests and responsibilities.

Decreased face-to-face social interactions.

Emotional outbursts, lying, and rule-breaking when limits are set on AI or screen use.

Sleep disturbances related to excessive nocturnal device use.

Sharing inappropriate personal information or images with AI systems.

Using AI to cheat on school assignments.

Using AI tools to bully peers and creating sexual or violent content with AI.

Ways to Help Youth Use AI Safely

The AACAP flyer includes recommendations that are shared by experts who study the influence of the internet on children and adolescents. Parents should:

Explore the AI tools together with their child, both to look for questionable content and become familiar with an AI platform’s features, such as parental controls and access to chatbot logs.

Model skepticism and how to adopt a questioning stance toward information offered by AI.

Encourage children to share and discuss online experiences and help them to understand that images, videos, audio, or depicted events may be fake.

Discuss the difference between a chatbot companion and human relationships, as well as promote and support human social interactions.

Set and enforce clear boundaries around AI and/or screen usage.

Learn and enforce school rules governing AI use on assignments.

Ensure that their child’s sensitive personal information or images are not being shared with AI systems.

The Problem With These Recommendations

The problem with these recommendations is that they put a great deal of the onus on parents, who are expected to stay up-to-date in a fast-moving, youth-oriented cyberworld that is outside of their experience and interest. Meanwhile, AI and social media providers remain broadly immune from legal liability under a 1996 law (CDA in 47 U.S.C. 230). This 20th-century law was originally intended to protect providers from accountability for libelous content created by others but posted on their platforms (Walker, 2023).

Thirty years later, the internet is a very different world. No longer composed of simple bulletin boards passively accepting user posts, AI-based websites are highly interactive, covertly, but profoundly, influencing users through sophisticated algorithms that are not well understood even by the providers.

In many respects, the situation has the feeling of a runaway train, and the nation’s youth are tied to the tracks ahead. It is past time for Congress and AI chatbot providers to meet their moral and societal obligations to actively deter dangerous AI-driven interactions, such as chatbots that enable suicidal impulses or risky behaviors, especially as they involve minors.

If you or someone you love is contemplating suicide, seek help immediately. For help 24/7, dial 988 for the 988 Suicide & Crisis Lifeline, or reach out to the Crisis Text Line by texting TALK to 741741. Outside of the U.S., visit the International Resources page for suicide hotlines in your country.

American Academy of Child and Adolescent Psychiatry (AACAP), AI and Children Committee, Is AI dangerous for children? Bulletin No. 145; July 2025.

Khushboo et al., 2026. Bullying and suicide attempts among US high school students. JAMA Network Open. 2026;9(1):e2552089. doi:10.1001/jamanetworkopen.2025.52089

McBain et al., 2025. Use of generative AI for mental health advice among US adolescents and young adults. JAMA Network Open. 2025;8(11):e2542281. doi:10.1001/jamanetworkopen.2025.42281

Pratt et al., 2024. Digital dialogue – How youth are interacting with Chatbots. JAMA Pediatrics May 2024 Volume 178:429.

Walker, J.T. (2023). Internet providers still enjoy broad immunity from liability for content. New Hampshire Business Review.

© Psychology Today