The Friend Who Can’t Refuse You

A lot of people are choosing AI companions over human friends right now, and it’s not just because they’re lonely or broken. It’s easy to look at this trend and write it off as something affecting only the isolated, the anxious, or the agoraphobic, whatever label helps the rest of us feel like this is someone else’s problem. The reality, however, is much messier. Some of these users have packed social calendars and active social lives, yet they still choose to open a chatbot app at the end of the day. They aren’t replacing human relationships; they’re choosing to use these apps alongside them.

That is the part that should make us uncomfortable. It means we can’t just comfort ourselves with the idea that “broken people are turning to broken substitutes.” It means these AI programs are offering something that is genuinely competitive with what we offer each other. Before we jump to a quick, comforting judgment, we need to take a closer look at what is actually happening.

The standard, lazy critique is that AI companionship is just “fake intimacy”, a cheap counterfeit of the real thing, dangerous because someone might mistake it for genuine connection. I argue that AI companions aren’t failing to be friends in the way we think they are. In fact, the places where they do fall short reveal a lot about what we’ve quietly stopped expecting from one another.

To understand why, it helps to look back twenty-four centuries to Aristotle, who did the heavy lifting of sorting out human relationships in his Nicomachean Ethics. He broke friendship down into three neat categories: friendships of utility (we’re useful to each other), friendships of pleasure (we enjoy each other’s company), and friendships of virtue, where we genuinely wish for the other person’s good, for their own sake, because of who they truly are. Only that third kind, he argued, is friendship in its truest, highest sense. The first two are just friendship-shaped arrangements that happen to feel good as long as they remain convenient.

If we are being completely honest, most of our adult friendships fit squarely into those first two categories. Consider the co-worker you grab lunch with, the old group chat that only stays active because you all used to live in the same town, or the friend you call to vent to because you know they’ll want to vent back later. None of this is “fake.” It’s friendship by any normal, modern definition. It’s also not the deep, soul-level virtue-friendship Aristotle was talking about. We just rarely admit that because it sounds cynical.

An AI companion, on the other hand, is refreshingly honest about its place in this hierarchy. It doesn’t pretend to care about your well-being independent of your interactions with it. It exists solely because you opened the app. If you close it and never return, it won’t notice, it won’t grieve, and it won’t send you a passive-aggressive text three weeks later asking where you went. It offers exactly as much friendship as a utility arrangement allows, and it never pretends to be anything more. What’s so provocative about AI isn’t that it’s deceitful; it’s that it might be the most honest utility-friendship many people have ever experienced, simply because no one is pretending otherwise.

The real breakdown happens somewhere unexpected. In the sixteenth century, Michel de Montaigne wrote about his bond with Étienne de La Boétie, describing a friendship so absolute and profound that he said it was easier to describe what it wasn’t than what it was. The most important part of Montaigne’s description isn’t the intensity of the bond, but its irreplaceability. He wasn’t saying his friend was just the best available option in his social circle. He was saying their friendship was singular. No other human being, no matter how similar or available, could have stepped into La Boétie’s shoes, because the connection was tied to that one specific man, not to a generic category of person who happened to act like him.

This is the exact arena where an AI companion cannot compete, not because of a temporary software limitation, but by design. The very things that make a companion app great at its job – its endless availability, infinite patience, and flawless responsiveness to your personal quirks – are the exact traits that make it entirely replaceable. It is built to be a substitute. If you lose your data, switch apps, or upgrade your phone, a functionally identical companion can be generated instantly, perhaps even slightly improved. Montaigne wouldn’t have found this tragic; he would have found it nonsensical, like mourning the loss of a specific pair of everyday socks. There was never a unique individual there to lose in the first place.

The philosopher Martin Buber offered an even sharper take on this dynamic. His famous distinction between “I-Thou” and “I-It” relationships isn’t about how an interaction makes you feel; it’s about whether the other party meets you as a whole, living presence, rather than as an object to be used or evaluated. The unsettling reality raised by contemporary thinkers is that an AI companion might be incapable of being a “Thou” not because it lacks sophistication, but because there is structurally no one on the other side to meet you. You can easily have an I-Thou-shaped experience, even when the other half of the conversation is strictly an “It.” Buber wouldn’t call that a flawed friendship; he’d likely say it isn’t a relationship at all, but rather an illusion camouflaged well enough to look like one.

Going a step further than the traditional philosophers: I don’t think most users actually care about these academic distinctions, and I don’t think that makes them foolish. If the subjective experience of feeling heard and understood feels exactly the same whether or not a real mind is on the other end, we run into a much deeper, more uncomfortable question. Is being truly met by another soul actually the core of what we want from companionship? Or do we just want the feeling of being met, and have simply been putting up with the complexities of real people because they were the only source available?

The philosopher Emmanuel Levinas would have deeply resisted this idea. His entire ethical framework was built on looking into the face of another human being – an encounter that demands something of you precisely because the other person is vulnerable, mortal, and completely outside your control. An AI companion demands nothing. You cannot wrong it. You can be cruel to it, ignore it for months, or delete it entirely, and nothing real has been hurt because there was never a capacity for suffering there to begin with.

For many people, that is the ultimate selling point: a relationship with absolutely no ethical weight, no risk of guilt, and no chance of hurting someone’s feelings. I suspect this is the true appeal of these apps, far more than any marketing copy lets on. It is worth pausing to consider how strange it is that “a relationship where I can cause no harm” has become a feature people are willing to pay a monthly subscription for.

This brings us to the deeper issue beneath the surface. Lars Svendsen, in his extensive research on loneliness, points out a fact that flips the standard narrative on its head: a lack of trust in other people predicts loneliness far more accurately than a simple lack of social contact. The core issue isn’t isolation; it’s trust. When we look at it through that lens, the sudden boom in AI companions looks less like a shortage of human company being patched by a robot, and more like a massive trust shortage. We are patching the gap with something engineered from the ground up to never require trust in the first place, because a chatbot can’t betray you, judge you, get bored of you, or walk away.

The real “companion problem” is much heavier than the cliché that “people are lonely, so they’re settling for machines.” The real worry is that we have successfully manufactured a frictionless substitute that removes the exact ingredients – risk, vulnerability, and the very real possibility of being let down – that genuine trust and friendship require to exist.

An AI companion isn’t just a poor imitation of a human friend. It is a vivid glimpse of what is left over when you engineer every single ounce of risk out of being known. The question we need to sit with isn’t whether these apps count as real friends. It’s why so many of us would prefer a version of intimacy that costs us nothing at all.