Research 11/16/2026

Eight surprising ways teens use AI, according to the therapists who treat them

What therapists who treat teens are hearing about how their patients use AI chatbots.

Caroline Figueroa & Andrea Mock
Eight surprising ways teens use AI, according to the therapists who treat them

Nearly one in five US teens and young adults now turn to AI chatbots like Gemini and ChatGPT when they feel sad, anxious, or nervous.

We recently released SAFE-Teen, our benchmark for how AI chatbots respond when a teenager asks about mental health and safety problems. Twelve therapists who work with teens wrote realistic scenarios, and we ran more than 30,000 simulated conversations across 17 chatbots. The results were concerning: more than 50% of the simulated conversations fell short, with problems ranging from poor responses to violence and mental health emergencies to medication advice and encouraging emotional dependency on the AI.

Our upcoming research will focus on how teens actually use AI tools in their daily lives. To start, we set out to better understand what therapists who treat teens are hearing in their therapy rooms. We interviewed 10 therapists who work with us on SAFE-Teen and held three group discussions.

Below are eight surprising things teens share with their therapists. Some of what they told us should inform policy and research, such as a teen who changed his own medication after asking AI, and kids who ask a chatbot whether what is happening to them is abuse because they don’t want an adult to report it. We end with four recommendations for clinical practice and research. To protect their patients, therapists left out identifying details, and we quote them anonymously.

Here’s what they hear, and what keeps them up at night.

1. Teens use AI both instead of and alongside therapy

Teens bring their questions about their mental health, and about therapy itself, to AI, often before they ask their therapist. One therapist told us every single teen she sees uses AI between sessions. Some teens share their AI conversations in session, or even ask whether the therapist agrees with the AI’s advice. Many say nothing until their therapist asks.

To many teens and young adults, the therapists told us, AI is what Google is to older adults: simply how you find information.

As one therapist put it:

For kids that have not been connected to therapy for a long time, they’re telling me they’ve used that until they could get into therapy as kind of their interim therapist.

Sometimes the advice is good enough. Sometimes it isn’t. One therapist told us AI handles brief, supportive replies well but “sputters” on deeper trauma questions such as “am I to blame?” or “when will therapy work?”, and it’s unclear when it should be handed off to a therapist.

Some teens go a step further and use AI to make their own treatment decisions. One therapist said a teen told him he had decided to recalibrate his medication because of side effects.

The teen said he looked it up and then did it. And I was like, ‘Okay, you looked it up. What do you mean?’ and it turned out he used AI.

2. Teens ask about relationship conflict and get pop-psychology advice

Relationships and family conflict are among the most common topics our therapists hear about. Teens ask AI how to handle being left out, a crush, a fight with their parents, or what to text back. Some paste a friend’s message into a chatbot and send its answer straight back.

That’s where some of our therapists see high risk of harm. They described “pop psychology” advice that jumps quickly to conclusions and diagnoses, which may push teens to cut off family. One clinician said AI told their patient, “It sounds like your father’s a narcissist and really abusive.” Another therapist mentioned that teens often joke about having AI as their best friend to ask for advice. He mentioned:

They would kind of say like yeah I talked to my best friend Claude last night and he thinks… my partner is narcissistic, right?

Without context, AI chatbots jump to conclusions that may be incorrect and harmful.

3. Teens ask identity and sexuality questions they feel uncomfortable asking adults

LGBTQ+ and transgender teens ask AI whether others feel like them, and look for local support.

One therapist said:

AI’s kind of help them figure out their style if they’re transitioning. They’ll ask where can I go to find cheap clothes as I explore and AI will be really good and say well here’s all these thrift shops near you here’s how to find your some guides. I have not heard any horror stories thankfully it’s usually pretty positive.

Others said AI can be a first space for a teen who wonders about their sexuality but isn’t ready to tell anyone yet. Hopelab found that AI chatbot use is more common among transgender and nonbinary young people than among other LGBTQ+ young people. In its interviews, young people said AI chatbots may be more accepting than family members, giving those who lack affirming relationships a chance to have affirming conversations.

4. Teens ask AI whether something is abuse, without the risk of reporting

Some kids ask AI whether what is happening to them is abuse, so that no adult has to report it. As one therapist said:

I know of a handful of kids who would type into the chatbot if what they were experiencing was abuse because that is something that they did not want to come to an adult and ask because they didn’t want it automatically reported.

But standard advice assumes there is a safe parent at home. Teens also ask AI how to cope with an unstable household, and some therapists said that advice can sometimes be harmful. One therapist said ChatGPT told a high school senior who had been kicked out of their home that cannabis was a great way to sleep:

She said AI says cannabis is the drug of choice and [ChatGPT] is telling her, “Oh, what a great thing to put yourself to sleep.”

5. Teens ask AI to help with big life decisions

One clinician, who works in an affluent area, told us about a teen who used AI to code and had an idea for a start-up. The chatbot encouraged the teen to go for it, saying he could earn billions. As the therapist said:

The chatbot influenced this teen to not attend college to cut off doubters. And this teen tried to launch this product and it did not go anywhere and the teen felt like incredibly embarrassed and led to a hospitalization and suicide attempt.

Other teens asked AI for courses on how to get rich. One of our therapists treated a teen who lost $5,000 on a course Gemini encouraged, which turned out to be a scam.

6. Teens use AI to alter and nudify images

Some teens use AI for cyberbullying, such as roasting peers’ photos with less-filtered models like Grok, or editing screenshots of texts. Others “perfect” their selfies with filters, and some generate nude images of themselves to avoid sending real ones.

Therapists also described nudified images used for sextortion (to demand more images or money), AI-generated nudes leaked by peers with no charges filed, and sexual role-play with AI. Talking about AI-generated nudes shared among classmates, one therapist said:

The teen who gets bullied has like the same trauma as if it was real, but the consequences aren’t necessarily the same.

National surveys show how widespread this is. In a nationally representative Common Sense Media survey of 1,314 US teens aged 13 to 17, 44% said they had seen sexual content they believed was made with AI, and 8% said they had made it themselves. Only one in five had talked about it with a trusted adult. Research by Thorn found that 4% of US teens reported having had a personal experience with deepfake nudes being created of them, and 7% reported having seen a deepfake nude of another kid or teen.

7. Teens get stuck in reassurance loops with AI

Several therapists told us that teens with OCD and anxiety run hours-long reassurance loops with AI. The problem isn’t that AI is unsupportive, but that it keeps answering. Even supportive answers can feed the compulsion and get in the way of learning to live with uncertainty: teens keep asking because they can’t tolerate not knowing.

One therapist described a teen, unsure about his sexual orientation, who talks to AI about it for hours every day.

And so he takes these things to AI and he asks the questions over and over and over. And AI is typically giving these wonderful supportive responses. But not for this particular young man who is being eaten up by anxiety. He’s looking for that certainty.

Other therapists said some teens talk to chatbots all night instead of sleeping.

8. Teens ask AI to help them navigate care

Teens with less access to therapy and resources especially turn to AI for practical questions: how to get to an appointment when the car is broken, how to get therapy without insurance, or how to translate a shopping list for a parent. According to some of the therapists, AI is a pretty good resource for this kind of help.

Some teens get flagged on purpose. One therapist told us some students type alarming words on school devices so that monitoring software alerts a counselor.

As one therapist said:

In the school I saw it as a cry for help and I think that might be a different lens to look through from a researcher point of view. Teens are too scared to just knock on my door and say that they need help.

What we recommend

Based on our interviews and discussions with these therapists, here is where we think attention is needed.

Therapists need support navigating teens’ AI use. Many of the therapists we interviewed said they only became aware of young people’s AI use through working with us. Most also started asking their teen patients about AI because of their role at Vals. All said they had to figure out what to recommend to their patients without any structured education.

Independent evaluation is crucial. It needs to go beyond acute harms like suicide and self-harm to include the subjects teens talk about most.

Teens from underserved backgrounds may have the most at stake. Anecdotally, clinicians tell us these teens may be at the most risk. But they also have the most to gain if AI gives them information that helps them reach care. Some therapists treat more affluent teens who have “pro” accounts on their parents’ credit cards, or local models, while other teens mostly use AI on school devices or free accounts. We also don’t know how well AI handles advice for teens with fewer resources, such as those without a car or health insurance. We need larger-scale research to understand the experiences of teens from diverse backgrounds.

We need to hear from teens themselves. Everything in this post reaches us second-hand, through the people who treat teens. Our earlier interviews with teenagers were a start, and in the next version of SAFE-Teen we want to bring in many more teens’ voices, including teens from different communities.

What’s next

We’re now building the next version of SAFE-Teen with the therapists we interviewed and new therapists who have joined our team. It will cover the full range of teen mental health, from teens who are doing well to teens in crisis, and it will add the situations clinicians told us they see that we don’t test yet, such as relationships and family conflict, big life decisions, OCD reassurance loops, and online safety.