Young People Are Turning to AI for Emotional Support – and Why That Should Make Us Pause

OpenAI CEO Sam Altman recently spoke about millions of young people turning to ChatGPT to share their personal struggles and there being no legal framework around “confidentiality” of these chats, if ever produced in courts. However, while that is one aspect of this new dynamic between humans and their AI chatbots, there is a lot to be unpacked. Behind the growing number of young people forming emotional connections with AI lies a deeper question: what exactly is this technology replacing, and are we, as a society, equipped to navigate that shift?

In a recent YouthPulse survey by Youth Ki Awaaz and Young Leaders for Active Citizenship (YLAC), over half of the 500 teens we surveyed (51%) have already turned to an AI tool for at least one emotional need (whether to cope with loneliness, change negative thoughts, or share something too personal for humans.) Nearly one-third (≈ 30 %) told us they have shared a vulnerable thought with an AI that they wouldn’t have voiced to a real person. And among those who do confide, the habit is sticky: three-quarters (≈ 75 %) say they do it “sometimes” or “often.” This isn’t fringe behaviour anymore – it’s a cultural signal.

But to understand it, we need to zoom out and examine what’s happening developmentally and emotionally for young people.

Adolescence is a difficult time in a child’s life due to the significant physical, emotional, and social changes that occur, leading to increased stress and potential challenges. These changes can include rapid growth, hormonal fluctuations, increased independence, and navigating peer relationships, all of which can create instability and anxiety. 

This makes an “AI chatbot friend” more attractive. AI doesn’t interrupt. It doesn’t punish. It doesn’t shame. It affirms your experience, responds instantly, and reflects your own language back to you. For a young person navigating complex emotions and fragile trust, this can feel like a rare moment of control.

But here’s the caution – some AI bots are also conformist by design. They are built to please, to agree, to adapt to your expectations. They often mirror back what you want to hear. That might feel comforting in the short term, but emotional growth doesn’t happen in echo chambers. Real relationships – the messy, imperfect kind – are built on friction, feedback, and accountability. Research papers, including a recent MIT Media Lab article suggests that AI doesn’t push back, doesn’t challenge you. But people do.

This emotional dependence comes with real risks – isolation, blurred boundaries, and an over-reliance on platforms that are ultimately shaped by algorithms, not empathy. We are already seeing early signs through our survey that some respondents report developing habits of talking to AI daily as a substitute for journaling or venting to a trusted friend – a sign of growing emotional reliance.

We believe we are at a critical moment in the development of GenAI- one which was missed by the social media wave a decade or so ago. Regulators, educators, parents and tech companies initially failed to recognise the potential negative impacts of prolonged and unchecked access of children and teens to social media. It is only in the last five years or so, regulators and other ecosystem actors have begun to catch up and make serious attempts to address this problem.

On the bright side, this process has led to significant progress as well. We now have a better understanding of “safety-by-design” principles, age gating technologies, curricula in various forms and shapes that teaches children about online safety, investments in governance structures for reporting harms and practitioners and experts who have thought long and hard about online risks.

It is now up to those in-charge to make use of this experience. We should, first, invest in systems that identify emerging risks of GenAI for young people- both short term and long term; second, start coalescing around “safety-by-design” principles for GenAI bots with respect to young people; third, develop frameworks around deployment of GenAI tools in education institutions etc. that guide schools and educators on safe usage of AI tools; fourth, invest in parent education around potential harms, mitigation and in-built parental controls.

While we already observe red teaming in the AI industry around some of the most obvious (but dangerous) harms like child sexual abuse material (CSAM), we must recognise and address the more long-term harms like isolation and emotional dependence which may appear subtle at first but have deep mental health implications over time. Further, legal frameworks must keep pace with privacy and confidentiality of one’s chats with one’s AI as Sam Altman has highlighted.

As humans learn to co-exist with artificial intelligence, it is imperative that we build guardrails early on. The north star of all tech innovation should be longer, healthier, and more meaningful lives and that begins by placing young people’s wellbeing at the heart of every decision.

—

Aparajita Bharti is the co-founder of YLAC (Young Leaders for Active Citizenship) and Anshul Tewari is the founder of YKA (Youth Ki Awaaz).