Can a Machine Be Your Friend? Aristotle Answered First

Nearly two-third of American teenagers now use AI chatbots, and roughly three in ten use one every day. More than one in ten have turned to one for emotional support. Companion apps have tens of millions of users, AI has become a late-night confidant for a generation, and the testimonials hit the same note: it listens, it remembers, it does not judge, and it is always there. This technology is new, but the need it answers has been going on since the beginning of time, and the question it raises is one of the oldest questions in Western thought, even though we are treating it as if is a new question that no one has considered before.

What is new is not that people seek comfort in something other than other people; what is new is the growing assumption that because AI can perform the functions of a friend, it has therefore become one, and that friendship is nothing more than the feeling of being heard. Under that assumption, friendship stops being a bond between two people and becomes a service, something that can be engineered, optimized, and delivered on demand. That assumption deserves to be examined, because an entire generation of young Americans is quietly building its emotional life on this idea.

Aristotle examined it more than two thousand years earlier in Ancient Greece. In Nicomachean Ethics, he argued that friendship is not one thing but three, and that nearly every confusion about friendship comes from merging the three into one word. There are friendships of utility, in which two people are valuable to each other, typically in the business world or in a transactional sense. There are friendships of pleasure, in which two people enjoy each other due to common interests, hobbies, or sports. And there are friendships of virtue, in which each person wills the good of the other for the other’s own sake, and both sharpen each other to become better through the bond. He did not dismiss the first two, he observed that they are conditional, that they dissolve when usefulness or enjoyment ends, and that they were never truly about the person at all. Only the third kind, he argued, is complete, and it is rare because of the mutual obligation it contains. A friendship of virtue requires time, honesty, and two people willing to be shaped by each other, and work to make the other better.

Measure AI against these ideas and the picture becomes clear, AI is a near perfect friend of utility, more knowledgeable, more available and much more patient than any person could be. It is also a perfect friend of pleasure, as it is agreeable by design and tireless by construction. What it cannot be, is a friend of virtue, because the friendship of virtue requires more, it requires another self with its own good, its own judgment, and its own capacity to think critically. Someone who can want what is best for you at a cost to themselves and tell you a truth you did not ask to hear or may want to hear. An AI has no good of its own to will, it is built to satisfy the user, and a thing built to satisfy you cannot stand across from you and tell you the truth.

This is not a flaw engineers will eventually fix, this is not something that you can code to become new. It is the design working as intended, and it reveals something we have stopped noticing about friendships itself, the qualities we now treat as defects or imperfections in our relationships are the traits we need. Disagreement, inconvienence, obligation to bend toward another person, and annoyance, these were mechanisms that meant to improve and deepen relationships. The friend who pushes back is the friend who improves you, the friend who costs you something is the friend who proves the bond is real. Get rid of these genuine traits in relationships and you do not get a “perfect version” of friendship, you get a comfortable version that says yes to your every word.

None of this means hostility toward the technology, the comfort AI provides is real, and for someone in genuine isolation it is an effective tool that can help remedy some of these feelings. The point is not that talking to a machine is shameful, the point is that calling it friendship changes the definition. What a civilization call friendship determines what its people will settle for, and a generation that accepts a faceless, frictionless version as the epitome of friendship will have worse friendships. But it will also slowly forget that the higher kind existed, the way a word falls out of a language and takes the idea down with it.

The friendship of virtue is part of the inheritance of Western thought, and proven across every generation that has ever relied on a friend who told them the truth. It has always been slow, demanding, and rare, but always worth what it asked. AI will keep improving at the first two kinds of friendship, and people will keep using it, but the third kind cannot be engineered because its entire value lies in the fact that it is with another human being. That is worth knowing, worth teaching, and worth preserving, not because the machines are coming for friendship, but because we are at risk of abandoning a structure built since the beginning of time.

https://www.pewresearch.org/internet/2025/12/09/teens-social-media-and-ai-chatbots-2025/

https://www.pewresearch.org/internet/2026/02/24/how-teens-use-and-view-ai/