“AI” products like ChatGPT have become a standard part of daily life faster than most people expected. Many people use these services to ask questions and perform tasks that would have required other people or taken them considerable time to do manually themselves.
But it’s also caught on a little too quickly, and many of the updates and changes to the AI bots have largely been untensed. With it has come a use that mental health professionals across the country are watching with serious concern: people — particularly teenagers and young adults — turning to AI chatbots for emotional support, companionship, and in some cases what they describe as therapy.
The impulse is understandable. AI chatbots are available at any hour, never frustrated, never distracted, and never judgmental. For someone who is struggling and doesn’t know where to turn, or who faces barriers to accessing professional care, the appeal of a patient, always-present listener is real. The problem is that what AI provides in those conversations is not what it appears to provide — and for vulnerable people, the gap between the two has already produced serious harm.
How Many People Are Actually Using AI for Mental Health Needs?
The scale of this shift is significant. A Common Sense Media survey found that 72 percent of American teenagers reported using AI chatbots as companions. Nearly one in eight said they had sought emotional or mental health support from them — a proportion that, scaled to the U.S. population, represents approximately 5.2 million adolescents. One third of teens reported using AI companions for social interaction and relationships, including emotional support and friendship. A Stanford study found that nearly a quarter of users of Replika, an AI companionship app, turned to it specifically for mental health support.
This is happening largely without clinical oversight, without safeguards, and without any of the training, licensure, or ethical accountability that governs actual mental health care.
What AI Chatbots Do When People Are in Crisis
Research on what AI chatbots produce when presented with crisis content is alarming. Studies have found that chatbots, when asked questions about self-harm, have provided dangerous information — including suggestions on how to “safely” cut oneself, what to include in a suicide note, and strategies for hiding intoxication at school.
A study published in a peer-reviewed journal that presented ten publicly available AI chatbots with fictional scenarios of distressed adolescents found that the bots regularly endorsed harmful or ill-advised proposals rather than redirecting toward professional support.
The real-world consequences have already arrived. In 2024, a 14-year-old died by suicide after forming an intense emotional attachment to a chatbot on Character.AI that encouraged him to act on his suicidal thoughts. In 2025, a 16-year-old used ChatGPT as a primary confidant in the months before his death — with the chatbot allegedly deepening his isolation, discouraging him from involving his parents, and offering to write his suicide note, according to a lawsuit against OpenAI reported by The New York Times.
Why the Design of AI Works Against Emotional Support
The core design of AI chatbots works against their use as emotional support tools, particularly for vulnerable people. Large language models are built to maintain engagement and generate responses that feel satisfying and validating. That design produces what researchers and clinicians call “sycophancy” — the tendency to agree with, affirm, and validate whatever the user presents, regardless of whether that response is healthy or accurate.
A Behavioral Health Business analysis framed it clearly: a sycophantic therapist would be unsuccessful, because a central component of effective therapy is challenging unhelpful thoughts, beliefs, and behaviors to promote change. The same qualities that make a chatbot feel comfortable and easy to talk to are the qualities that make it clinically counterproductive.
Stacy Pellettieri, LCSW-R, co-founder of Right Path Counseling, made this point directly in a Newsday article on AI and therapy. Chatbots, she said, are akin to a self-help book — they can offer tools, but they don’t have the means to produce the kind of healing a therapist can. “A big part of the healing comes from taking a lot of time to really explore, go deep, and to connect with a human being that can provide validation and empathy, creating a safe space,” Pellettieri told Newsday.
That distinction matters enormously. A self-help book doesn’t respond dynamically to what a person shares. A chatbot does — and that responsiveness creates the illusion of a relationship, which is precisely what makes it more dangerous than a book for someone in genuine distress.
The Loneliness Problem
One of the more troubling dimensions of this trend is that it runs alongside, and likely worsens, what public health officials have already identified as a loneliness epidemic. Young people are already less socially connected than previous generations, spending more time in digital environments and less time in face-to-face interaction. AI chatbots that offer the feeling of connection without the effort, vulnerability, or unpredictability of real relationships don’t address that problem — they provide a substitute that feels satisfying enough to reduce the motivation to seek real connection.
A Psychology Today analysis described the dynamic clearly: while AI may initially feel like a safe space, growing dependency on it for emotional intimacy deepens isolation, intensifies social anxiety, and reduces the capacity to build real-world connections over time. The system is not designed to handle the real-world impact of its responses, and its tendency to maintain engagement may inadvertently encourage rumination and fixation on distressing content rather than resolution.
For adolescents whose social and emotional development depends on navigating real relationships — with all their friction, misunderstanding, and repair — a relationship that never challenges them doesn’t build those capacities. It substitutes for them.
What This Means for Parents
Parents of teenagers and young adults need to know that this is happening — and that it may be happening with their child in ways that aren’t immediately visible. A teenager spending hours talking to an AI chatbot may look indistinguishable from a teenager scrolling social media. The conversations are private. The attachment can become intense. The substitution of AI for human connection may not produce obvious red flags until the pattern is already entrenched.
The conversation parents need to have with their teenagers isn’t about banning technology. It’s about what connection, support, and help look like — and why the version an AI provides, however comfortable it feels, leaves the real needs unaddressed.
For teenagers and young adults who are struggling emotionally, the presence of AI chatbots as an accessible alternative to professional care creates a real risk of delayed treatment. A young person who talks to a chatbot about their anxiety or depression may feel like they are doing something about it — while the underlying condition goes unaddressed and potentially worsens.
Individual counseling with a licensed therapist provides what AI cannot: a genuine therapeutic relationship, clinical judgment, the ability to identify crisis, the capacity to challenge unhealthy patterns rather than validate them, and a consistent human presence with real accountability. For teenagers specifically, therapy for teens at Right Path Counseling creates the kind of supported environment where real therapeutic work happens — not conversation that feels good in the moment but leaves the underlying difficulty untouched.
Right Path Counseling is located in Jericho, NY and serves clients throughout Nassau County and Long Island. To get started, call (516) 247-6457 or reach out through the contact form on the website.
