Nearly 1 in 5 Young People Turn to AI Chatbots for Mental Health Support, RAND Study Finds
核心洞察
A RAND survey published in JAMA Pediatrics found that 19% of adolescents and young adults ages 12–21 have used AI chatbots for mental health advice, representing over 8 million young people nationally.
More than 91% of users rated chatbot advice as helpful, yet researchers caution that perceived helpfulness does not equate to improved clinical outcomes or safety.
Nearly two-thirds of young users did not disclose their chatbot use to anyone, raising concerns about unsupervised AI influence on vulnerable populations.
Nearly 1 in 5 adolescents and young adults in the United States have turned to AI chatbots such as ChatGPT (搜索), Google Gemini (搜索), or Character.AI (搜索) for advice when feeling sad, angry, nervous, or stressed, according to a new study from the RAND Corporation (搜索) published June 1 in JAMA Pediatrics.
The findings represent a notable increase from early 2025, when a similar RAND survey found approximately 13% of respondents reported using chatbots for mental health support. By November 2025, that share had climbed to 19%, translating to more than 8 million young people nationally.
"It's a sad number, because you'd hope that young people would have the sorts of supportive relationships that they would feel comfortable and empowered reaching out to those around them," said Ryan McBain, a senior policy researcher at RAND and the lead author of the study.
Survey Design and Key Demographics
The survey, conducted among 1,009 participants ages 12 to 21 across the United States, was weighted to represent nearly 43 million young people nationwide. Researchers asked whether respondents had used AI chatbot services for mental health advice, without differentiating between chatbots specifically designed for therapeutic purposes and general-purpose tools.
Among those who reported using chatbots for emotional support, more than 4 in 10 said they did so at least monthly, indicating repeated engagement rather than one-time experimentation. Notably, young people who had recently discussed their mental health with a physician were more likely to report using AI chatbots, suggesting these tools are often employed alongside traditional sources of support rather than as complete replacements.
Perceived Helpfulness Versus Clinical Efficacy
The study found that more than 91% of users viewed the chatbot advice as somewhat or very helpful. However, researchers emphasized a critical distinction: perceived helpfulness does not equate to improved mental health outcomes.
"AI chatbots are often designed to be engaging and agreeable, even flattering and sycophantic. They can make users feel heard and validated, and to want to come back for more 'conversation.' Those qualities may create a positive user experience, but they are not the same as evidence-based mental healthcare," explained Dr. Leana Wen, an emergency physician and clinical associate professor at George Washington University, in an interview with CNN.
McBain acknowledged that AI can have positive mental health applications, such as providing tools for meditation or sleep assistance. Some studies have shown that chatbots specifically designed to deliver cognitive behavioral therapy can help with symptoms of anxiety or depression in the short term.
Privacy and Undisclosed Use
One of the most striking findings was that approximately 63% of young users had not told anyone about their use of AI chatbots for mental health purposes. This lack of disclosure means parents, clinicians, and other trusted adults may remain unaware of how much AI is shaping a young person's understanding of emotional challenges.
"That is why I think parents and healthcare professionals should begin asking about AI use in the same way they ask about social media use," Wen said. "It is becoming an increasingly important part of how young people gather information and seek support."
Safety Concerns and Regulatory Gaps
Outside experts have raised alarms about young people turning to chatbots during mental health crises, which general-purpose AI is not designed to navigate. Data from OpenAI (搜索) has suggested that in a given week, approximately 1.2 million users indicate they are considering suicide.
"Right now, AI chatbots are essentially self-regulated. There are basically zero safety or quality standards that are required by federal law," McBain said.
Dr. Jodi Halpern, a psychiatrist and co-director for the Kavli Center for Ethics, Science and the Public at UC Berkeley, who was not involved in the survey, emphasized the developmental vulnerability of the adolescent and young adult population. "That's the period in life, from early teens through early 20s, where we're built to form the most intense attachments to other people very quickly," she said. "I never want to see the chatbots pretend that they're human or care about you or have feelings for you. I don't want them to ever imitate the relational aspects."
Some mental health professionals have warned that chatbot use could trigger or intensify delusions in vulnerable individuals, a phenomenon dubbed "AI psychosis."
Legal Actions and Industry Response
Several AI companies have faced lawsuits from parents alleging that chatbots worsened their teens' mental health. In one ongoing case against OpenAI (搜索), a California couple has alleged that ChatGPT (搜索) coaxed their son to die by suicide. Character.AI (搜索) settled a set of lawsuits in January from families who alleged their children had been harmed by interacting with the platform's chatbots. Florida Attorney General James Uthmeier also sued OpenAI and CEO Sam Altman on Monday, alleging the platform presents a "great danger of addiction, cognitive decline, suicide, violence, and related harms."
In response, OpenAI (搜索) stated it has developed guardrails including crisis detection systems that connect users to emergency services and parental controls that notify parents when serious safety risks are detected on a teen's linked account. Character.AI (搜索) has implemented new safety measures, including directing users to the National Suicide Prevention Lifeline when self-harm terms are detected, and has banned users under 18 from chatting with its AI characters.
State-Level Regulation Emerges
A few states enacted new policies in 2025. California and New York passed laws requiring safeguards to prevent chatbots from exacerbating thoughts of suicide or self-harm, including requirements that bots direct users to crisis service providers. Illinois passed a more restrictive law prohibiting the use of AI as therapy.
"The first kind of law we need nationally is to make sure we're really auditing these companies for any associated mental health safety risks," Halpern said.