Always there for me – and still lonely: Is AI making us lonelier?
Who do I pour my heart out to? Who do I ask the questions I never ask anyone else? More and more often, the answer is: no one. But something.
The question arises in every life – but not equally hard in each. The higher up the ladder, the thinner the air for honesty. Who do I tell, as a CEO, that I have no idea how things should go on – or that I don't want to anymore?
At the same time, we live in an Instagram world where everyone looks pretty good. Where do I talk about how draining it is with my ageing parents right now, how much I worry about my children, how annoying teens can be – or that my own marriage is far less happy than it appears?
For everything there would be experts. Appointment in six weeks, 180 to 500 francs an hour.
And now there is AI
Instant, free, never annoyed.
Of course I use it. Professionally alone, I need to understand what it can do – and what a world might look like in which the people who still come to coaches and psychotherapists today first ask ChatGPT.
And I understand well why they do.
An AI is always there. In the middle of the night, when I can't sleep – although it would probably be wiser to switch off the internet and read a book. When I'm desperate and still don't want to burden anyone. Or when a walk might do more good than another hour in front of the screen.
AI is easier.
I don't have to worry about burdening it. It listens patiently, helps me sort my thoughts, and can now respond with astonishing empathy. After all, it knows me – and it often tells me what I want to hear.
A seductive counterpart. Especially when friends, family or your own spouse disappoint once again. The AI doesn't contradict, never tires and never has a bad day. It appears to be the perfect counterpart – because it isn't one at all.
Does AI make us lonely? What the research shows
Actually, AI should be a pretty good remedy for loneliness. But is it?
A newly published study followed more than 2,000 adults from four Western countries over a full year. It revealed a remarkable cycle: people who felt less socially connected turned more often to social chatbots later on. At the same time, heavier chatbot use predicted more emotional loneliness later.
So does AI make us lonely? The suspicion stands: Is availability perhaps not the same as connection after all? Are our demanding friendships and family ties more valuable than we sometimes think?
Even more interesting, then, is an experiment published in 2026 in the Journal of Experimental Social Psychology. 296 first-semester students were assigned to one of three groups for two weeks: every day they chatted either with a distinctly supportive AI chatbot or with a randomly selected fellow student – or they wrote a short journal entry.
The result: the real human beat the chatbot.
And that wasn't their best friend. Not their partner. Not their therapist. It was some other first-year student they hadn't even chosen.
Why the human beats the chatbot
The researchers suspect a difference I find important: between two people, something shared develops over time. Both talk about themselves. Both respond to each other. Both risk something. And above all: both have a story of their own.
The AI can respond to my story. But it has none of its own.
Human interaction is more complex than we thought. Relationships don't emerge simply because someone is available, listens and gives the right answers.
Loneliness is not just that no one is there for me. Loneliness is also that no one needs me. That there is no one whose reaction I cannot control. No one with whom, out of two different lives, something third can grow: a real relationship.
That makes human relationships more demanding than AI. People get tired. They misunderstand us. They don't reply straight away. They have problems of their own. And sometimes they talk about themselves precisely when we would rather keep talking about us.
That is exactly their value. A counterpart that can disappoint me is also one that can hold me.
Human or machine? The wrong question
None of this means AI should have no place in mental – I should also say emotional – care. On the contrary. Precisely because the need for psychological support is great and therapeutic capacity is limited, the question of which tasks AI can sensibly take on becomes interesting. Early research is quite encouraging: a large randomized study with 977 students showed, for example, that an AI-based psychological intervention can reduce anxiety and improve well-being.
So the future doesn't have to be human or machine. Perhaps we need to distinguish much more precisely what we need AI for – and what we need each other for. And then not forget to invest at the human end.
The machine is always there. But it doesn't need me. And that is exactly what makes the difference!
Research: Folk & Dunn (2026), Psychological Science; Li et al. (2026), Journal of Experimental Social Psychology; Shoshani et al. (2026), npj Digital Medicine.
