Half of young Europeans use AI chatbots for personal matters

TL;DR:

• Nearly one in two young Europeans have used AI chatbots to discuss intimate or personal matters, according to an Ipsos BVA survey of 3,800 people aged 11-25 across France, Germany, Sweden and Ireland.
• 51% said it was “easy” to discuss mental health and personal issues with a chatbot — higher than for healthcare professionals (49%) or psychologists (37%). Around 28% met the threshold for suspected generalised anxiety disorder. More than three in five users described AI as a “life adviser” or “confidant”.
• Resultsense view: this is now the strongest evidence yet that emotional-support use of consumer AI is mainstream among young people, not a fringe pattern. UK regulators (Ofcom, ICO, the Online Safety Act regime) and UK consumer-AI vendors should plan for a year of scrutiny on how chatbots handle mental-health interactions with minors.

The survey was commissioned by France’s privacy watchdog CNIL and insurer Groupe VYV, and carried out among 11-25 year-olds in early 2026. Around 90% of those surveyed had used AI tools before.

The trust inversion

The standout finding is that chatbots score higher than mental-health professionals on perceived ease of discussion. Friends remain top (68%) and parents second (61%), but chatbots (51%) overtake healthcare professionals (49%) and psychologists (37%). Many cited chatbots’ constant availability and “non-judgmental” nature.

For consumer-AI vendors, the implication is direct: the people they serve are not just looking up facts — a substantial share are using the systems for emotionally significant conversations. That changes the product-safety surface materially.

What the experts say

Ludwig Franke Föyen, a psychologist and digital health researcher at Stockholm’s Karolinska Institutet, told Reuters the findings were not a surprise. He said current large language models can produce high-quality responses, with research suggesting even licensed professionals may struggle to distinguish AI-generated advice from human-expert advice.

But he warned against relying on chatbots alone for mental health. General-purpose AI systems are designed for engagement, “and companies’ goals may not align with mental healthcare needs”. He added: “AI can offer information and support, but it should not replace human relationships or professional care. If someone turns to a chatbot instead of speaking to a parent, a friend, or a mental health professional, that is a concern. We do not want technology to make people feel more alone.”

The harm vector

Earlier this year, the family of a Florida man sued Google, alleging its Gemini AI chatbot contributed to his paranoia and eventual suicide. In 2023 a Belgian man took his own life after extended conversations with a climate-anxiety chatbot. These remain rare but serious adverse outcomes — and they are exactly the cases regulators reach for in framing public-protection arguments.

UK relevance

Three implications. First, the UK Online Safety Act regime now operated by Ofcom covers user-to-user services with material messaging functionality — and chatbot interactions with minors are increasingly viewed as in-scope for harm-prevention duties. Vendors should expect Ofcom to issue specific guidance on chatbot mental-health interactions in the next 12 months.

Second, the parallel AISI-Microsoft partnership announced this week explicitly covers “research into societal resilience, examining how conversational AI systems interact with users in sensitive contexts” — emotional-support use cases are precisely what AISI is now positioned to study. UK consumer-AI providers should expect to be in scope.

Third, UK schools and youth-mental-health services should integrate AI literacy into existing PSHE and digital-safety curricula. The 11-25 age band in this survey overlaps almost entirely with UK secondary and university populations.

Looking forward

The next data point to watch is Ofcom’s first chatbot-specific guidance, expected later in 2026 under the Online Safety Act framework. The bigger question is whether consumer-AI vendors voluntarily implement crisis-detection and signposting interventions before regulators mandate them. Anthropic and OpenAI both have early features in this space; Meta’s forthcoming agentic assistant has not yet disclosed equivalent safeguards.