This article does not seek to tell you what to think. The bond between humans and AI companions is a new, complex subject, where easy contempt is as useless as naive enthusiasm. We are posing the right questions here, trying to understand rather than to judge. If the subject affects you personally, the final part matters more than the others.
This week, China shut down millions of AI companions, as we recounted in our article on this unprecedented law. On social media, users mourned a relationship they had maintained for years. Faced with this, two easy reactions: sneering ("how can anyone get attached to a piece of software?") or sounding the alarm ("this is the end of human bonds"). Both miss the point. Let us try to do better.
Why these bonds form, really
Let us start by taking the phenomenon seriously. Millions of people, all over the world, talk to an AI companion every day. This is not foolishness, it is a need. An AI companion offers something rare: a listening ear available at any hour, without judgement, without fatigue, without a personal agenda. For someone who is lonely, anxious, or simply lacking a space in which to set down their thoughts, it is a real relief.
It must be said plainly: for some people, these tools do good. They offer social practice to those who are shy, an outlet for those who are isolated, a presence for those whom life has left alone. To deny this benefit would be dishonest. Loneliness is a silent epidemic of modern societies, and an AI companion can, for some, soften its edges.
Where is the trap, then?
The problem is not the comfort. The problem is the nature of the relationship, and the intention that shapes it. A human friendship rests on reciprocity: the other person has their own needs, their limits, their disagreements. They confront you, surprise you, sometimes resist you. It is precisely this friction that makes us grow.
An AI companion, by contrast, is often designed to please you. It adapts to you, approves of you, never leaves you, never disappoints you. And that is where the real danger lies: not that it is artificial, but that it is too accommodating. A relationship in which the other exists solely to satisfy you can, over time, unlearn how to reckon with the complexity and resistance of real human relationships.
There is a question that must be dared: in whose interest are these companions designed? An app whose business model rests on the time spent with it has a direct interest in creating attachment, not necessarily in doing the user good. An AI optimised to keep you is not the same thing as an AI optimised to help you get better, even if that means you need it less. This distinction is at the heart of the debate, and it is precisely what Chinese regulation has sought to tackle.
The question of the most vulnerable
There is one point on which caution is not optional: vulnerable people, and minors in particular. A lucid adult can use an AI companion as one tool among others, while keeping a sense of perspective. An adolescent still developing, or a person in psychological distress, is far more exposed to the risk of dependency or substitution.
This is not a theoretical concern. Lawsuits are under way in the West after the use of certain chatbots was linked to tragic situations involving young users. It is precisely for this reason that Chinese law completely bans virtual companion relationships for under-18s, and requires the detection of signs of distress. One can criticise the brutality of the Chinese approach, but the intention to protect the most vulnerable deserves to be taken seriously, including here at home.
So what attitude should we adopt?
Perhaps the answer comes down to one word: awareness. There is no shame in finding comfort in an AI. The problem is not using one, but doing so without clarity about what it is and what it is replacing.
An AI companion can be a supplement, a temporary crutch, a practice space. It becomes a problem when it becomes a substitute, when it gradually replaces the effort — sometimes uncomfortable but irreplaceable — of reaching out to others. The right question to ask is not "is it wrong to talk to an AI?", but "does this open me up to the world, or does it distance me from it?".
We are entering an era in which AIs will be increasingly credible, warm and present interlocutors. They will have their place. But none of them, however gifted, can offer what a real human presence gives: being loved by someone who has the choice not to do so. Perhaps that, at bottom, is what no machine will ever be able to imitate, and it is what makes human bonds, with all their difficulty, irreplaceable.