Artificial intelligence is becoming a familiar source of emotional and mental health advice for young people in the United States. A new national study found that almost one in five people aged 12 to 21 had used an AI chatbot for mental health guidance, marking a sharp increase from the previous year.

The findings suggest that AI is already becoming part of how younger users seek support. Researchers say the most urgent question is no longer whether this is happening, but whether the advice young people receive is accurate, appropriate and safe.

Use Has Risen Rapidly

The study, published in JAMA Pediatrics, surveyed a nationally representative group of young people in the US. Of 1,727 individuals contacted, 1,009 responded, with results weighted to represent more than 42 million people aged 12 to 21.

Almost 20% reported using AI chatbots for mental health advice. Earlier research had placed the figure at roughly one in eight, showing how quickly the behaviour is spreading.

Older participants were more likely to use AI for mental health questions, and use was also more common among young women. Researchers found no significant difference between metropolitan and non-metropolitan areas.

Most Users Tell No One

One of the study’s most important findings involved disclosure.

Most young people who used AI for mental health advice had not told a parent, clinician or another trusted person. That matters because many general-purpose chatbots were not originally designed to provide clinical mental health care.

Hao Yu, associate professor of population medicine at Harvard Medical School and one of the study authors, said this lack of disclosure should encourage parents and healthcare professionals to ask young people directly about AI use rather than assuming they would mention it themselves. He also stressed that trusted people may help users distinguish reliable information from misleading or inappropriate advice.

Why AI Is So Appealing

The attraction is easy to understand. Chatbots are available at any hour, require no appointment and can respond immediately.

That convenience is particularly important in a healthcare system facing shortages of mental health professionals, including specialists who work with children and teenagers. AI may therefore appear to fill a gap when professional support is difficult to access.

It can also feel less intimidating to type a private question into a chatbot than to begin a conversation with another person.

But accessibility does not automatically make the information clinically reliable.

Useful Does Not Always Mean Safe

More than 90% of young users in the survey said they considered the AI advice they received useful. Researchers caution, however, that perceived usefulness is different from proven effectiveness.

A chatbot can sound reassuring and confident even when its response is incomplete or inappropriate. Many AI systems are also designed to maintain engaging, agreeable conversations, which can become problematic when a user needs challenge, escalation or professional assessment rather than simple reassurance.

Yu noted that researchers still need much stronger evidence about the safety and clinical effectiveness of mental health guidance produced by AI systems. He argued that evaluation should focus on whether these tools meet appropriate clinical standards, not simply whether users enjoy interacting with them.

AI Cannot Replace Professional Care

The findings do not show that AI chatbots are equivalent to therapy, diagnosis or treatment.

General-purpose systems may offer basic information, help users organise thoughts or encourage them to seek support, but they can miss context, misunderstand risk or provide inaccurate information.

The challenge is especially serious when conversations involve severe distress, self-harm or other urgent mental health concerns. Researchers say stronger safeguards, independent testing and clearer standards are needed as use continues to grow.

What Comes Next

The research team expects AI use for mental health questions to continue increasing and plans follow-up surveys.

That makes oversight increasingly important. Rather than trying to ignore or prohibit a behaviour that is already widespread, experts are focusing on improving safety, transparency and communication.