More and more young adults are using AI chatbots to talk about stress, anxieties, or personal problems. This is shown not only by our own research, but also by large European studies. But what does this mean for mental health? And how can we use AI sensibly and safely? Here we highlight both opportunities and risks and explain why digital competence and critical thinking are so important right now.
High usage, but little confidence
For an Erasmus+ application, we surveyed 43 young adults between 18 and 25 from Austria and Romania. The results are striking: 95% know generative AI tools like ChatGPT, 75% actively use text-based AI, and more than half use AI-driven social media features. Particularly striking is that 30% of respondents regularly use AI for emotional support when they are stressed or overwhelmed. At the same time, many feel unsure: only 17% are confident that they can check AI information for accuracy. Almost half describe the influence of AI on their wellbeing as mixed or unclear. Many name feeling overwhelmed by the flood of information, comparison with AI-generated images, and doubts about the accuracy of the content. These data are of course not representative. But they show trends in our target group – and match what larger studies in Europe also observe.
European trends: almost one in two talks to AI about personal matters
A recent survey of 3,800 young people between 11 and 25 from France, Germany, Sweden, and Ireland shows similar patterns. Almost half of the respondents use AI to talk about personal or intimate topics. 51% find it easier to talk to a chatbot about mental health than to psychologists (only 37%). A third even see AI as a therapist in some situations. In Italy, too, 42% of respondents between 13 and 19 said they use AI when they are sad, feel anxious, or need help with important decisions. These figures underline that for many young people, AI has long been more than a tool. It is becoming an advisor, sometimes even a substitute for human conversation.
Where AI can really help
Using AI is not inherently bad for our mental health. AI-supported offerings hold a lot of potential – especially where there are gaps in care. AI chatbots are available around the clock, anonymous, and often free of charge. They can reach people who do not seek therapy because of shame or high costs. Studies show small to moderate effects in the reduction of depression, anxiety, and stress through AI chatbots, especially when they are based on cognitive behavioural therapy. They can help sort feelings, guide breathing exercises, or provide information about mental health. AI can also serve as a complementary tool between therapy sessions. It can help track progress or set reminders for self-care routines.
The risks of AI
Despite the opportunities, experts warn urgently against overestimating AI in mental health care. AI cannot read body language, cannot make clinical diagnoses, and often cannot respond appropriately in acute crises. Studies have documented cases in which chatbots reinforced negative thought patterns or responded indifferently to suicidal thoughts. AI programmes that claim to have their own personality and a friendship with their users in particular can be dangerous for young people. In numerous firsthand accounts, affected people describe how they developed unhealthy, almost dependent attachments to such personalised chatbots.
Why AI cannot replace therapy
Many users stay with AI systems instead of seeking professional help – even though the systems can be ineffective or even harmful and have been shown to reinforce delusions. A Stanford study warns: AI can amplify bias and stigmatisation and discourage people from seeking further help. What you confide in a chatbot is not subject to confidentiality. Many systems are not developed for therapeutic purposes, and clear rules and quality standards are missing.
AI carries our societal biases
Initiatives like AI Empower, which work on AI ethics, point out: AI is not a neutral mirror. It carries exactly the risks and biases that we already have in society – from gender stereotypes to racist patterns. This means: when AI talks about mental health, it can also amplify prejudices, make wrong diagnoses, or disadvantage certain groups. For marginalised young people in particular, this carries a high risk. UNESCO therefore emphasises: AI systems must be developed transparently, fairly, and under human supervision – especially when they work with vulnerable groups.
What young people want and need
AI is neither good nor bad. It is a tool – and as with every tool, what matters is how we use it. For young people, this means: AI should be a complementary support, not a replacement for therapy or human conversation. It is important to critically question what AI says and always check whether the sources are trustworthy. It is also important to set boundaries and to know when it is time to seek professional help.
Young people don't just want to use AI – they also want to understand it. The majority of respondents in our survey want to know how AI works. They are interested in its effects on mental health. They want to learn how to protect their data and recognise AI-generated fake content.
This is exactly where we come in. We know that AI can be a valuable tool – but only if people have the competence to use it critically and self-determinedly. It takes digital resilience to protect yourself from overload, manipulation, and false information. And it takes human connection, because no AI can replace real empathy.
From September 2026 to June 2027, we are therefore running an Erasmus+ project focused on AI literacy, critical thinking, and peer education – co-created with and for young adults aged 18 to 25, who will create their own campaigns over the course of the project.
Are you curious and would like to take part? Then get in touch at josephine@stabilimkopf.com