As AI tools slip quietly into our routines, a growing number of people are talking to ChatGPT as if it were a friend. A recent study from OpenAI and the MIT Media Lab offers a closer look at how these interactions really shape daily life — and the findings raise both fascinating and troubling questions about our emotional dependence on technology.
A sense of closeness that can become complicated
The study analysed 40 million chatbot conversations, surveyed 4,000 users, and followed 1,000 participants over a month. One clear pattern emerged: those who talk to ChatGPT every day often develop a stronger emotional attachment than those who use it occasionally. Researchers describe this as a form of emotional dependence, a digital closeness that can slowly blur the line between companionship and habit.
I’ve heard similar stories from friends who work remotely and turn to AI during long, quiet afternoons. It starts innocently — asking for recipe ideas, help drafting emails, or a sounding board during stressful moments — but the convenience can gradually feel like comfort. According to the study, this effect is especially pronounced among younger adults, who tend to spend more time in virtual environments.
How the voice mode makes the relationship feel more human
One of the most striking insights concerns ChatGPT’s voice mode. Interacting with a lifelike voice makes the experience feel more personal, even intimate. When used moderately, it can actually reduce feelings of loneliness. But when relied on too heavily, it increases the likelihood of forming addictive patterns.
The researchers note that spoken conversations trigger different emotional cues than text does. A warm tone or a natural pause can create a sense of human-like presence that text doesn’t offer. Yet the paradox is apparent: while voice chat can make people feel less alone, it can also make the attachment feel more real than it actually is.
The strange balance between openness and distance
Another unexpected discovery concerns how people use the AI. Users who share personal stories or vulnerabilities tend to feel lonelier afterward, even though they’re opening up. Meanwhile, those who stick to lighter, more neutral topics often develop stronger dependency — but without feeling closer in any emotional sense.
This dual effect raises tricky questions. If discussing personal issues increases loneliness, and avoiding them increases dependence, where exactly should the boundaries lie? Psychologists interviewed for the research point out that this dynamic echoes patterns seen in other forms of digital interaction, where the illusion of connection can mask a more profound sense of isolation.
A growing concern for the future of human–AI relationships
For now, the researchers stress that intense emotional attachment to AI is still relatively rare. But the indicators are there, and they’re becoming harder to ignore. As conversational AI becomes more natural, more accessible and more present in our daily tasks, the ethical challenges grow too.
Designers and policymakers may soon need to decide how far these systems should go in mirroring human behaviour. Should an AI sound comforting? Should it acknowledge emotions? And how can developers prevent users from slipping into digital solitude, where a machine becomes their primary confidant?
These questions aren’t about limiting technology but about ensuring it coexists with human well-being. After all, AI can be a remarkable tool — many people credit it with boosting productivity, creativity and confidence. But as this research makes clear, the line between support and substitution is thin, and navigating it will be one of the defining challenges of our increasingly connected world.
English (US) · 
