Teens Using AI for Emotional Support: 12 Questions Counselors Can Ask Without Shaming Them

“I asked AI about it.” A teen might say this while describing a fight with a friend, a difficult night, or something they have not felt comfortable telling anyone else. They may mention it casually. They may also watch your face to see whether sharing it was a mistake.

For therapists and school counselors, that moment opens a conversation about teens using AI for emotional support: what this young person needs, what they are getting from a chatbot, and where they may need more support. You can take the risks seriously while still making room for an honest answer.

Here are 12 questions to help you begin, along with ways to discuss accuracy, privacy, emotional reliance, and human connection.

AI chatbots and teen mental health: what new research says

Two reports released on October 27, 2026, give practitioners a timely reason to bring AI into conversations about young people’s lives.

Common Sense Media’s 2026 Census found that 49% of tweens ages 9 to 12 had used an AI chatbot or companion at some point. That figure describes use broadly. It does not mean that half of tweens are using AI for emotional support.

The Jed Foundation’s new research examined responses from more than 5,500 middle and high school students across 21 U.S. schools participating in JED’s programs. In that sample, 57% reported using AI when feeling sad, stressed, lonely, or overwhelmed. Other reported uses included relationship advice, problem-solving, and companionship.

JED found associations between certain uses of AI and poorer mental health and well-being, including greater histories of suicidal thoughts and attempts. These findings deserve attention, but they do not show that AI caused those outcomes. As JED explains in its discussion of the study’s limitations, the data are cross-sectional and the sample is not nationally representative. Young people who are already struggling may be more likely to turn to AI, and the study cannot determine which came first.

For an individual session, these findings give you a reason to ask. They cannot tell you what one particular teen’s AI use means.

Start with curiosity

Before offering information about AI, give the teen room to describe their own experience. If they bring it up, you might begin:

“I’m glad you told me. I’d like to understand what those conversations are like for you.”

If you introduce the topic yourself, use language that leaves room for a no:

“Some people use AI to talk through personal things. Has that been part of your experience?”

The American Psychological Association’s guidance for practitioners discussing AI with patients recommends asking about use, exploring what feels helpful or unhelpful, and inviting continued conversation.

You do not need to work through all 12 questions. Choose the ones that fit the young person’s age, communication style, and reason for being there. These are conversation prompts, not a validated assessment or a substitute for clinical judgment. If a safety concern emerges, respond to it right away rather than finishing the list.

A note from Danielle Lloyd, Certified Coach: I’m a coach, not a therapist. I share these questions as conversation starters from my own work with kids, teens, and young adults. If you are a licensed clinician, use your own clinical judgment and follow the standards of your profession and setting.

Understand how the teen uses AI

1. “Which app or chatbot do you use, and how would you describe talking with it?”

Start by learning what “AI” means to this teen. They might be using a general chatbot, a fictional character, or a product marketed as a companion.

Invite them to describe the experience in their own words. Do they think of it as a search tool, an advice giver, a character, or someone who knows them?

You do not need to know every product to have a useful conversation, and it is best not to guess at its features or safeguards. A teen can explain what they experience without opening their account or handing over their phone.

2. “What kinds of things do you usually talk through with it?”

Keep this broad at first. Homework help, jokes, relationship advice, identity questions, and emotional reassurance may all come up.

If they struggle to answer, offer a few possibilities without making one sound like the expected response: “School stuff, things with friends, feelings, something else?”

Listen for the role the conversation plays. A teen practicing how to ask a teacher for help and a teen relying on a chatbot to decide whether a friendship is worth keeping need different follow-up conversations.

3. “What makes it easier to ask AI than another person?”

This question makes room to discuss barriers to human support. A teen might describe embarrassment, fear of a lecture, difficulty finding words, or not knowing who would be available.

Resist the urge to immediately list all the people they could have asked. First, understand what feels hard about those options.

If they say, “It doesn’t judge me,” you might respond: “Feeling judged makes it hard to talk. What do people do that gives you that feeling?”

Their answer may also tell you something useful about how to make your own sessions feel more approachable.

4. “When you open it, what are you hoping it will do?”

Are they looking for an answer, reassurance, a distraction, words for a message, or a place to put a feeling?

Try reflecting that need back in plain language: “You wanted somewhere to say it before you were ready to tell a person.”

Acknowledging the need does not mean endorsing the chatbot’s response. It helps you identify what a safer, more reliable source of support would need to provide.

Explore how the conversations affect them

5. “How do you usually feel afterward?”

Make room for mixed answers. A conversation might feel comforting in the moment and leave the teen more worried later. They may also feel disappointed, confused, or unchanged.

You can invite them to walk through one recent example: how they felt before opening the app, right afterward, and later that evening.

A report of feeling better is not proof that the interaction was safe, and a report of feeling worse is not proof of its cause. Either way, their experience gives you something specific to explore.

6. “Do you ever find yourself asking the same thing again because the answer doesn’t stay reassuring?”

This can open a conversation about repeated checking and difficulty tolerating uncertainty.

Ask what happens between the first answer and the next question. Does the relief last? Does a new doubt appear? Does the teen feel able to stop?

The APA’s health advisory on chatbots and mental health warns that these tools may reinforce reassurance-seeking and rumination for some people, including those with anxiety or OCD. Explore the pattern within the teen’s broader clinical picture rather than assuming that repeated questions point to a diagnosis.

7. “Has it ever said something that felt wrong, uncomfortable, or too intense?”

Give the teen permission to question a response even if they liked other parts of the conversation.

They might describe advice that ignored important details, a response that escalated a conflict, or language that felt unusually personal. Ask what they noticed and what they did next.

If they describe dangerous advice, respond clearly and calmly: “That advice could put you at risk. I’m glad you brought it here.” Then assess the safety implications without blaming them for receiving or believing the response.

Discuss accuracy, privacy, and boundaries

8. “How do you decide whether its advice is worth following?”

Explore what makes a response seem trustworthy. Does it sound confident? Agree with them? Offer a detailed explanation? Remember something they said before?

You might offer a made-up example: “Imagine a chatbot says you should end a friendship after hearing about one argument. What would it need to know before giving that advice?”

The APA’s adolescent well-being advisory highlights the importance of helping young people question AI-generated information and tell simulated empathy apart from human understanding.

A response can sound caring and confident while still missing something important or being wrong.

9. “What do you think happens to the things you type into it?”

Treat this as a shared information gap, not a test they should already have passed.

Explain that a conversation that feels private is not necessarily confidential. Depending on the service and its settings, messages may be stored, reviewed, shared, or used to improve the system. Avoid making promises about a specific product without checking its current policies.

Talk through why names, school details, addresses, identifiable screenshots, and sensitive information about other people deserve caution. Privacy settings may reduce some risks, but they do not make a consumer chatbot equivalent to a confidential clinical relationship. The APA’s practitioner guidance specifically recommends discussing these limits.

10. “Does talking to it ever get in the way of sleep, school, or time with people?”

Ask about everyday effects without assuming there is a problem. A quick exchange and a conversation that regularly runs into the early morning have different implications.

You can also explore whether the teen feels increasingly dependent on an AI response before making decisions, or whether they are pulling away from friends and family they once saw often.

JED’s guidance for parents and caregivers encourages adults to stay curious while watching for increased reliance, withdrawal, and distress when AI is unavailable. These signs are reasons to look more closely, not a stand-alone diagnosis.

Make human support easier to reach

11. “Who feels safe enough to go to when you need more support?”

Allow an honest “I don’t know.” Do not assume that every parent, caregiver, or other adult is a safe option for every young person.

Explore what “safe enough” would look like. Perhaps they need someone who will listen before offering advice, take them seriously, or explain what happens after a disclosure.

Support may include an appropriate caregiver, school counselor, therapist, relative, or another trusted adult. Friends can offer connection, but serious distress and safety concerns call for adult and professional support.

If no one comes to mind, helping them identify and reach appropriate support becomes part of the work.

12. “What would make it easier to reach a person the next time things feel hard?”

Turn a general suggestion into a concrete next step. Depending on the situation, that might mean rehearsing a sentence, learning how to request a school counseling appointment, or identifying an appropriate adult who is available outside school hours.

For everyday support, a simple opening message could be: “I’m having a rough day. Can you listen for a few minutes?”

For safety concerns, use a direct, individualized plan with qualified human support and emergency options. In the U.S., that plan can include the 988 Suicide & Crisis Lifeline, which teens can call or text at 988. Make clear that a chatbot should not be relied on to assess danger or provide crisis care. The American Counseling Association’s AI recommendations explicitly caution against relying on AI for crisis response.

When a disclosure raises a safety concern

Conversations about AI sometimes reveal an immediate concern: unsafe advice, self-harm or suicidal thoughts, threats, abuse, exploitation, or a young person who feels unable to keep themselves safe.

At that point, move from general discussion into your established safeguarding and clinical response procedures. Ask directly about the specific safety concern, assess current risk within your competence, and arrange appropriate human support. Do not rely on the chatbot’s reassurance or conclusions about the young person’s safety. The APA warns that chatbot crisis responses can be unreliable.

Follow the professional, organizational, and legal requirements that apply to your role and setting, including emergency response, consultation, documentation, confidentiality, caregiver involvement, and mandated reporting. Explain the limits of confidentiality in language the young person can understand. When possible and safe, involve them in planning what happens next.

Avoid automatically asking for a full chat history. Consider whether specific content is needed for assessment, and how any review, collection, or storage of it will be handled under your setting’s privacy and consent requirements.

Create room for connection in the session

A teen may understand the limits of AI and still find talking to a person hard. Give them chances to experience a conversation where they have time to think, can disagree, and do not need a polished answer.

Sometimes that starts with a shared activity. A short game, a drawing prompt, or a quick check-in gives you both something to respond to before moving into more personal territory. Ask permission before switching activities, and leave room to keep talking if that is what the teen prefers. For more ideas, see our guides to therapy activities for teens and therapy check-in questions.

That is the idea behind Opening Play, a tool for leading sessions with kids, teens, and young adults. A game like Plot Twist Doodle or a Session Enhancement like The IDK Menu gives you and the teen something to work on together, with you guiding the conversation. It works in person or over telehealth by sharing your screen (see our telehealth games), there is nothing to download, and clients don’t need accounts. In keeping with everything above about privacy, client names, answers, and drawings stay in the browser and are never saved to Opening Play’s servers, and no AI analyzes client answers.

When a teen tells you they have been talking to AI, you have a chance to understand what has felt helpful, notice what may be unsafe, and make human support easier to reach. Start by making it possible for them to keep telling you the truth.

Editor’s note: This article has not been reviewed by a licensed clinician. It offers general information, not clinical guidance, so please follow your own professional standards around safety, confidentiality, and caregiver involvement. Research figures come from reports released October 27, 2026; check the original sources for the most current findings.