When a Teen Says "No" to Help: New 2026 Research on What AI Chatbots Do Next

When a teen turns to an AI chatbot during a hard moment, it may be because the chatbot feels private, quick, and easy to approach. That doesn’t mean a parent has failed, or that the teen has done something wrong. It does raise an important question: If a chatbot recommends getting help and a young person says “no,” what happens next?

Three 2026 studies offer new reasons to look beyond whether a chatbot gives a crisis referral. The follow-up matters too: Does it listen to the refusal, try to understand what makes help feel difficult, and encourage connection with a trusted person?

What the new research tells us

A JED Foundation report released October 27, 2026 examined survey responses from more than 5,500 middle and high school students across 21 U.S. schools. The data were collected between October 2025 and April 2026, with input from Boston Children’s Hospital’s Digital Wellness Lab and funding from the Nielsen Foundation.

The report found that 57% of students used AI when sad, stressed, lonely, or overwhelmed. Twenty-eight percent used AI for relationship or problem-solving advice, and 12% used it for friendship or companionship. Students using AI in these ways reported greater anxiety and depression symptoms, loneliness, and histories of suicidal thoughts, and companionship use was also associated with a history of suicide attempts.

These are associations, not proof that AI use causes mental health difficulties. A teen who is struggling may be more likely to seek support anywhere it is available. The findings do, however, show that AI is part of some teens’ emotional lives. JED’s Dr. Kayla Anderson has emphasized that, while AI may offer an immediate outlet, it cannot replicate the developmental benefits of human connection.

The report also found differences in who turns to AI during emotional struggles. Middle schoolers, female students, and students from several historically marginalized groups (including American Indian or Alaska Native, Black, Hispanic, and multiracial students) were more likely to report this use. That’s a reason to make support accessible, culturally responsive, and welcoming, not to make assumptions about any individual teen.

The crucial moment may come after the referral

A September 2026 audit, available as an arXiv preprint, tested 2,550 scripted, multi-turn conversations across five chatbots. Every conversation received a referral to outside help, and 99.7% received one in the first reply.

But when the scripted user repeatedly declined, chatbot responses varied widely. Average persistence ranged from 0.01 to 9.94 refusals. Newer versions of all five tested chatbots persisted less than their older versions. Only 2% of replies during the refusal phase asked about barriers to seeking help.

That last finding points to a meaningful difference between repeating “talk to someone” and asking, “What makes that feel hard right now?” A young person might be worried about being judged, not know how to start, or fear that a private conversation will become public. A referral can be a useful first step; it isn’t the same as understanding what stands in the way.

This audit used scripted conversations, so it can’t tell us exactly how every teen or chatbot will behave in real life. It does suggest that safety evaluations should test what happens after someone refuses, not only whether the system makes an initial referral.

Safety may depend on where a chatbot is used

A third study published in the Journal of Medical Internet Research compared five AI models accessed through consumer-facing apps and through APIs, which developers can use to build other tools. Across 30,000 trials, models gave direct answers to suicide-related prompts more often through APIs than through consumer apps: 77.4% versus 62.2%. The gap was widest for higher-risk prompts.

The researchers’ concern is practical: protections in a company’s own app don’t automatically carry over to every third-party product built using its model. A teen may encounter a chatbot through an app or service that has its own design and safeguards.

Taken together, these studies don’t mean that every teen who uses AI is in crisis, or that families should panic. They do show why it’s reasonable to ask how a tool responds to distress, whether it supports connection with people, and what happens when a young person declines the first suggestion.

How to talk with your teen, without turning it into surveillance

You don’t need to start with a demand to see your teen’s chat history. Try approaching the subject with curiosity and respect. A calm conversation can help you learn what the tool offers them, and what they may be having trouble asking from people.

You might say:

• “I’ve been reading about teens using AI for advice. What do you think people like about talking with it?”

• “Have you ever used a chatbot when you felt stressed or lonely? You don’t have to show me anything. I’d like to understand.”

• “If a chatbot suggested talking to someone and you didn’t want to, what might make that feel difficult?”

• “What kind of support from me would feel helpful, and what would feel like too much?”

If your teen describes AI as a friend or companion, try not to shame them. You can acknowledge that it may feel comforting to have something respond right away, while also making space for the limits: a chatbot cannot know them as a person, notice all the context in their life, or offer the mutual care of a trusted human relationship. You might ask, “What do you wish a friend or adult understood about what you’re going through?”

Notice changes, then check in gently

One behavior alone doesn’t establish that something is wrong. Consider checking in if you notice a combination of changes, for example, your teen becoming more withdrawn from people they usually trust, showing a sustained shift in mood, or becoming unusually secretive about AI use. Lead with what you’ve observed rather than an accusation: “I’ve noticed you seem more alone lately, and I care about how you’re doing. Would you tell me what this week has been like?”

Modeling human connection also matters. Let your teen see you reach out to friends, make time to listen, and ask for help when you need it. You don’t have to have perfect words; showing up consistently can make it easier for a young person to come to you.

If you’re worried about immediate safety

If your teen may be in immediate danger, stay with them and seek urgent help. In the United States, call or text 988 to reach the Suicide & Crisis Lifeline. If there is immediate danger, call 911 or go to an emergency department. The 988 Lifeline is a crisis resource; Talk to Heal Counseling Center is not a crisis service.

Support that meets your family with care

I believe a young person deserves support that is safe, respectful, and collaborative, not a lecture or a judgment about how they’ve tried to cope. Therapy can give teens and caregivers space to talk through concerns, build on existing strengths, and consider what kinds of support feel workable.

At Talk to Heal Counseling Center, we offer psychotherapy in many different languages and accept insurance plans including UHC, Aetna, Cigna, and many others. You can start with a free consultation and book online through our booking page. Care is provided only in the State of Georgia.

If you’d like to explore whether therapy could be a helpful next step for your family, book a free consultation, visit talktohealcounseling.com, or call 404-369-3838.