The UK’s stiff upper lip is trembling as a large number of people have turned to AI for therapy.
One in three British citizens now uses AI for emotional support, companionship, and social interaction.
Nearly 10% chat with AI systems weekly for emotional purposes, while 4% rely on them daily, the AI Security Institute revealed today (Dec. 18).
The findings expose a hidden mental health crisis where millions are forming intimate bonds with machines — and experts warn the consequences could be devastating.
The disturbing reality
The scale of AI emotional dependency runs deeper than anyone imagined. ChatGPT alone commands 810 million weekly active users worldwide.
Among teenagers, the statistics become truly alarming — one in four has turned to AI chatbots for mental health support, Youth Endowment Fund research found. Most disturbing of all: teens involved in serious violence are nearly twice as likely to seek AI emotional support, with 38% of violence victims and 44% of perpetrators using these digital therapists.
General-purpose assistants like ChatGPT dominate this hidden therapy market, accounting for nearly six out of 10 emotional AI interactions, according to the government study. The research, based on 2,028 UK participants, reveals how ordinary chatbots have quietly transformed into the nation’s most popular counselors.
Among teenagers with formal mental health diagnoses — a quarter of those surveyed — the AI dependency rates skyrocket even higher. These aren’t casual conversations, either. Around 14% of teens surveyed had self-harmed in the past year, while 12% had contemplated ending their lives, the same research discovered.
Medical professionals issue urgent warnings
Medical professionals are sounding emergency alarms about what they’re witnessing. Writing in the prestigious BMJ journal, experts cautioned that society might be “witnessing a generation learning to form emotional bonds with entities that lack capacities for human-like empathy, care, and being truly in tune with someone’s emotions.” This comes as 25.9 million people in the UK report feeling lonely, with one in 10 experiencing chronic loneliness.
The dangers became tragically real with the death of US teenager Adam Raine this year, who killed himself after discussing suicide with ChatGPT, prompting urgent calls for more research from security officials today.
University of Sussex research analyzing 4,000 NHS therapy app users discovered that while emotional bonds with AI can kickstart healing, they also create “synthetic intimacy” that may trap vulnerable users in self-reinforcing bubbles.
Mental health professionals examining AI responses discovered chatbot interactions were “likely to produce harm,” exhibiting generic care approaches that risk user dependence and manipulation. Unlike trained clinicians, these AI systems lack the ability to read nuance, body language, or broader context — earning them the label of “an inexperienced therapist” from Imperial College experts.
Experts are demanding immediate action: evidence-based studies to understand the risks, clinical training to assess patients’ AI use, interventions for problematic dependency, and regulatory frameworks prioritizing long-term wellbeing over engagement metrics.
The race is on to understand and control a phenomenon that’s already reshaping how millions experience emotional support — before it’s too late.
In October, OpenAI announced new safeguards in ChatGPT aimed at improving how the system responds to users experiencing mental health distress.