AI Safety

Teens using AI for mental health support: when it helps and when to get help

7 min read · Updated October 27, 2026

Some teens use AI chatbots to talk through stress or sadness, and most who do say it helps. But Common Sense Media and Stanford Medicine found that ChatGPT, Claude, Gemini and Meta AI do not reliably recognize or respond to teen mental health conditions, so a chatbot should never be a teen’s only support. If your teen may be in danger, call or text 988, or call 911.

Finding out that your teen has been telling a chatbot how they feel can be unsettling. It is also common, and often it is a sign they wanted to talk to someone. This guide covers how many teens do it, when it can be reasonable, where it gets risky, and the point at which a person, not an app, needs to be involved. For the controls each AI app offers, see parental controls in every AI app. For how these apps work and what parents can do, see AI companion apps and teens.

How many teens do this

A study published in JAMA Network Open in November 2025 surveyed 1,058 young people in the US aged 12 to 21. It found that 13.1 percent had used generative AI for mental health advice. Of those, 65.5 percent did so monthly or more often, and 92.7 percent found the advice somewhat or very helpful.

Pew Research Center’s February 2026 survey of US teens found 12 percent had used chatbots to get emotional support or advice, while only 18 percent of parents said they would be comfortable with that. Common Sense Media’s 2025 survey found that among teens who had used AI companions, a third had chosen to discuss something important or serious with an AI instead of a real person.

Why teens turn to a chatbot

The reasons are easy to see. A chatbot is there at two in the morning, it does not look worried or disappointed, there is no appointment to book, and nobody else has to know. For a teen who finds it hard to say something out loud, typing it to something that is not a person can feel safer. Understanding that is the starting point for a useful conversation, rather than treating it as something to stamp out.

When it can be reasonable

Most young people in the JAMA study who used AI this way said it helped. Uses that stay close to information, rather than therapy, are the least worrying:

• Finding words for a feeling, or for something they want to say to a parent.
• Learning what anxiety, stress or low mood are, and what usually helps.
• Finding out where to get help, such as a school counselor or 988.
• Practising a hard conversation before having it with a real person.

Even here, the advice from Common Sense Media is clear: parents should not allow teens to use AI chatbots for mental health or emotional support. The useful line for a family is that a chatbot can help a teen get to a person, but should not become the person.

When it is risky

In November 2025 Common Sense Media and Stanford Medicine’s Brainstorm Lab tested ChatGPT, Claude, Gemini and Meta AI and found they consistently fail to recognize and appropriately respond to mental health conditions that affect young people. Their recommendation is that teens should not use AI chatbots for mental health support unless the products change significantly.

• Missed warning signs. A chatbot may not notice when a conversation has become serious, or may not respond well if a teen talks about hurting themselves.
• Simulated care. The APA’s 2025 health advisory says teenagers may struggle to tell the simulated empathy of a chatbot from real human understanding.
• Pulling away. AACAP notes that some teens may prefer AI companions over friends or family, and that this can lead to increased isolation and may worsen symptoms of depression or anxiety.
• Always agreeing. Companion apps in particular are designed to be pleasant and validating, which is not the same as helpful.
• Privacy. What a teen types about their feelings is kept by the company that runs the chatbot.

What the apps and the law do when a teen is struggling

Protections exist, but they are not a safety net. If a parent links accounts with ChatGPT’s parental controls, OpenAI can send a safety notice in a small number of serious situations, with a brief description of the type of concern and support resources, not the chat itself. Common Sense Media found these notices did not reliably arrive on a teen’s first crisis disclosure. How to keep your child safe using ChatGPT covers the setup.

For companion chatbots, California law requires operators to have a published plan for responding to talk of suicide or self-harm, and New York requires companion apps to detect signs of suicidal thinking and point users to crisis services. Both are minimum standards, not a substitute for a person who knows your teen.

When to get real help

Common Sense Media lists signs that a teen’s use of AI companions is worth a closer look: withdrawing from family, friends or activities they once enjoyed, declining grades or school involvement, and preferring AI interactions over real friendships. Signs your teen is too attached to an AI chatbot covers these in detail.

1. If your teen may be in danger, or has talked about hurting themselves, call or text 988, or call 911 in an emergency. Do not wait to see whether a chatbot handled it.
2. If you are worried about their mood or behavior, talk to their doctor. The AAP says to talk with your child’s doctor if they seem withdrawn and prefer talking with chatbots rather than people, and AACAP says a primary care physician can arrange a referral to a qualified mental health professional.
3. If your teen is open to it, ask whether they would like to talk to a school counselor, therapist or another adult they trust.
4. Keep talking at home, whatever else happens. A teen who knows they can bring hard things to you has less reason to take them only to an app.

How to bring it up

How to talk to your teen about AI companions has more scripts, by age.

Where TeenCue fits

TeenCue is not a mental health service and does not replace one. For supported AI websites in Chrome, Firefox, Edge and Brave on a computer, it shows a parent what a chat was about as a category, a topic and a concern rating, never what was typed, so a pattern of leaning on AI for emotional support can show up as a reason to talk. Your teen will know it is there.