A new study has found that nearly one in five young adults and young adults seek advice from an AI chatbot when they are sad, angry, nervous or stressed.
Subscribe to read this story without ads
Get unlimited access to ad-free articles and exclusive content.
The findings, published by research organization RAND, show an increase compared to when the nonprofit organization conducted a similar study in early 2025. At the time, about 13% of respondents said they used chatbots for such advice, but that number rose to 19% in the group’s most recent survey in November, which was published Monday in JAMA Pediatrics.
“This is a sad number because young people would have wanted to build supportive relationships where they felt comfortable and empowered to reach out to those around them,” said Ryan McBain, senior policy fellow at Randland and lead author of the study.
In a new study, McBain and his team asked people between the ages of 12 and 21 whether they had used services such as ChatGPT, Google Gemini, and Character.AI for mental health advice. The survey questions did not differentiate between chatbots designed specifically to provide treatment and chatbots used for many purposes.
The researchers also asked whether the young people surveyed found the chatbot’s advice helpful, and the majority said it was. Approximately 63% of respondents said they had not told anyone that they were using artificial intelligence for treatment.
The proportion of young people using AI chatbots for mental health advice is similar to the proportion of young people who report receiving mental health therapy from a professional. The researchers said that while some people may be using both tools, they suspect many people are using AI chatbots as an alternative due to a lack of or lack of access to licensed mental health professionals. Other young people may use AI in these situations because they are already accustomed to using AI for other purposes.
But outside experts are concerned about young people. We turn to chatbots during mental health crises, but AI isn’t designed to help us deal with them. According to data from OpenAI, the developer of ChatGPT, 1.2 million users consider suicide in any given week.
McBain said the findings suggest further regulation is needed to ensure young people can use chatbots appropriately.
“Right now, AI chatbots are essentially self-regulated. There are basically zero safety or quality standards required by federal law,” he said.
However, McBain added that there could be positive uses for AI related to mental health, such as finding tools to help with meditation and sleep. Some studies have shown that in the short term, chatbots specifically designed to provide cognitive behavioral therapy (an approach that helps people identify unhelpful thinking patterns and change their behavior accordingly) can alleviate symptoms such as anxiety and depression.
Some people using AI chatbots for mental health purposes are taking part in online forums such as Reddit’s r/TherapyGPT. The forum maintains an active community of 28,000 visitors each week. Commenters have been exchanging tips on how to confide in chatbots, with some users describing them as a “lifeline.” One user uploaded a diary to ChatGPT and claimed it helped him sober up.
“Given me better advice than a real therapist. Told me what I needed to hear, not what I wanted to hear,” another user wrote. “Most of the answers made me think seriously about my life and even shed a few tears. I felt like I had found a breakthrough. Maybe it was because I was hungry for a real human connection with someone and couldn’t find it.”
But others on the subreddit have warned that chatbots are designed to over-verify, a warning also highlighted in a new study.
Some mental health experts say the use of chatbots can trigger or enhance delusions in vulnerable people, a scenario known as “AI psychosis.”
Other experts are concerned about young people developing parasocial relationships with chatbots.
“The period of life from our early teens to our early 20s is when we are primed to quickly form our strongest attachments to others,” says psychiatrist Jodi Halpern, Ph.D., co-director of the Kavli Center for Ethics, Science, and Public Affairs at the University of California, Berkeley. Not participating in this study.
“We never want to see a chatbot pretend to be human, care about you, have feelings for you. We never want it to imitate aspects of human relationships,” Halpern said.
Last year, some users suffered after OpenAI made changes that made bots less appealing. CEO Sam Altman said in a statement at the time that while it might be good to “use ChatGPT as a kind of therapist or life coach,” “if a user is in a mentally fragile state and prone to delusions, we don’t want the AI to enhance that.”
Some AI companies are facing lawsuits from parents who say their chatbots exacerbate mental health issues in teens. In one ongoing lawsuit against OpenAI, a California couple claimed that ChatGPT induced their son to commit suicide. Florida Attorney General James Usmeyer also sued OpenAI and Altman on Monday, claiming the platform poses “substantial risks of addiction, cognitive decline, suicide, violence, and related harm” to its users.
In response to questions about the California lawsuit, an OpenAI spokesperson said the company has been developing guardrails for its users over the years. These include a crisis detection system that connects people to emergency services, and parental controls that notify parents if a serious safety risk is detected on their teen’s linked account. OpenAI did not immediately respond to a request for comment on the Florida lawsuit.
When it comes to regulating the use of AI in mental health, several states have enacted new policies in the last year. California and New York have passed laws requiring safeguards to prevent chatbots from exacerbating thoughts of suicide or self-harm. For example, the requirement for chatbots to direct users to crisis management service providers. Illinois passed a more restrictive law banning the use of AI as a treatment.
“The first piece of legislation we need nationally is to make sure we’re actually auditing these companies for the associated mental health safety risks,” Halpern said.
