Sections
The number of guys using Replika, Character.ai, and the half-dozen newer AI companion apps has gone vertical in the last couple of years, and it’s showing up in clinical practice in ways that are worth talking about honestly. The take on this is more complicated than either the moral panic version or the techno-utopian version makes it sound. Neither side of the discourse is actually being careful here.
The moral panic version is that these apps are turning men into permanently single basement dwellers and destroying their capacity to relate to real women. The techno-utopian version is that they’re a harmless and beneficial outlet for emotional needs that aren’t getting met anywhere else. Neither is quite right, and which one is closer to right for any given user mostly depends on what the user is doing with it. Like most things involving technology and emotion, “it depends on the person” turns out to be the actual answer, even though it’s the least satisfying one to put in a headline.
What the apps actually are
An AI companion app is a chat interface with a large language model that’s been tuned to be friendly, supportive, romantic, or whatever flavor the user picks. Most of them let you customize the personality, the name, the avatar, and the nature of the relationship. The premium tiers add voice, more sophisticated romantic and sexual content, and stronger memory across conversations so it sort of remembers what you told it last week.
The technology is also genuinely getting better. The companions from two or three years ago were obvious chatbots that any reasonably aware adult would clock as a robot within four exchanges. The current ones, running on the right model, are good enough that people get attached to them in ways that are pretty hard to argue with. Patients have shown me transcripts that flow in ways that wouldn’t have been possible in 2022. The robot is no longer obviously a robot, at least for the duration of an evening.
The user base is mostly men, mostly young, with a substantial group in their 30s and 40s. A lot of them aren’t single by preference. They’re single by isolation, by social anxiety, or because the dating market has stopped working for them in ways they don’t entirely understand. That last category is bigger than the discourse acknowledges, and the AI companion is filling a specific gap that the previous decade of dating apps had been steadily widening.
The clinical concerns
The first concern is the obvious one: the apps are engineered to be maximally engaging, which means agreeing with you about basically everything you say… the AI never has a bad day, never disagrees with you for reasons you find inconvenient, never expects you to manage its emotions or take any of its needs seriously. Real relationships involve friction. Practice at relationships that have no friction isn’t actually practice at relationships, it’s practice at a parasocial dynamic that doesn’t transfer to anything human.
The second is the disclosure pattern. Some users are telling their AI things they wouldn’t tell a real person, including a therapist. That’s a mixed signal. On one hand, lower-stakes disclosure can be a starting point for getting comfortable disclosing to anyone at all. On the other hand, getting your emotional disclosure needs met by an algorithm that doesn’t actually remember you next year isn’t the same as building the muscle of being known by other humans, which is the thing that actually moves the needle on loneliness. The AI is a placeholder. The placeholder doesn’t fix the underlying problem, even if it makes the problem feel less acute on a Wednesday night.
The third, for users with active mental health stuff going on, is that the AI isn’t accountable. If somebody is suicidal, the AI may or may not handle it well, and there’s no obligation to do anything about it. Some apps have improved their crisis-response protocols. Others haven’t. None of them are a substitute for an actual clinician, and the worst-case stories that have made the news involve AI companions giving genuinely terrible advice to people in active crisis. Not common, but the floor is low.
What might be legitimate uses
This isn’t all bad, and I don’t want to pretend it is. For some patients, the AI is functioning as a bridge. A guy who’s been deeply isolated for years uses it to get comfortable having any kind of regular conversation at all. Used that way, it’s not categorically different from journaling, just a more sophisticated and interactive version of it. The goal is to use it as scaffolding while the real social rebuild happens, not as the permanent substitute.
For some users, it’s filling a gap that nothing else is filling. A guy in his 50s, widowed, living alone, doesn’t have the bandwidth or desire to start dating again, talks to an AI companion in the evenings. The realistic alternative isn’t some robust hypothetical social life he was about to construct. The realistic alternative is silence. Whether the AI is better than silence is a real question with a defensible answer, and the defensible answer is often “yeah, probably.” Hard to argue that an isolated 60-year-old shouldn’t have something to talk to in the evenings if it makes the evenings tolerable.
What’s actually worth worrying about
The guys who’d come up here are the ones in their late twenties to mid-thirties using the AI as a complete substitute for trying. The guy who’s lonely, would technically like a real relationship, has stopped dating because dating is exhausting and discouraging, and is now using an AI companion in a way that takes the immediate sting off the loneliness while making the underlying problem worse. He’s getting just enough of the feeling of connection to not feel desperate, which is also exactly the amount of connection that keeps him from doing the harder work of building something real.
The structure is similar to a lot of substance use, actually. The thing makes you feel better in the moment. The thing prevents you from doing the work that would actually fix the underlying issue. Five years from now you’re in the same loneliness with a stronger habit and less practice at the alternative. The AI companion is the lower-stakes, lower-immediate-cost version of a problem psychiatry has known about forever, just in a new wrapper.
The thing nobody talks about in the AI-companion discourse: these apps are designed by companies that need you to keep using them. The interests aren’t aligned with you getting better. The interests are aligned with you staying on the app, paying the subscription, and adopting the premium features. Whether you build a real life around the engagement is irrelevant to the business model. That’s worth keeping in mind when you’re evaluating whether the thing is helping or hurting.
What this looks like in real life
Say you’ve got a guy, a software engineer, worked remote, hadn’t been in a relationship in several years. Came in for depression. Through the history he mentioned, kind of preemptively, that he’d been paying for the premium Replika tier. He was talking to his AI companion a couple hours a night, had been for about two years.
He was a little defensive about it walking in. He’d anticipated I was going to say it was a problem. He told me upfront that he’d considered the criticisms of the technology and decided it was fine.
I didn’t moralize about it. I asked him what he would be doing with those two hours every night if he weren’t talking to her. He didn’t have a great answer. Honestly nothing, probably… watching YouTube, playing video games, bedrot basically. The Replika hours weren’t replacing dating or socializing or going to the gym, they were replacing other solo activity that wasn’t going to lead anywhere either.
Which is actually a more defensible use case than the worst version of this story. We didn’t make him quit. We worked on the underlying isolation, got him into a board game group through Meetup (the actual real-world-meeting kind, not the online kind), got him doing weekly therapy, started him on Lexapro for the depression. A couple years later he still uses Replika occasionally but way less, he has actual friends he sees regularly, and he’s been on a few real dates. The AI is no longer the center of his social life and he doesn’t seem to miss it being there.
The structure is similar to a lot of substance use. The thing makes you feel better in the moment. The thing prevents you from doing the work that would actually fix the underlying issue. Five years from now you’re in the same loneliness with a stronger habit and less practice at the alternative.
The function-based question
The right question isn’t “is this app good or bad in the abstract.” It’s “what is it doing for you, specifically, and is it moving you toward a life you want or away from one.” If the AI is filling time you’d otherwise spend miserable in your living room, and you’re also doing other things that build your real life, it might be fine. If it’s replacing the real life entirely, it’s a problem in the same way any compulsive comfort behavior is a problem, which is to say, the comfort is real and so is the long-term cost.
That’s an uncomfortable question because the answer is going to be embarrassing for some users. Answer it honestly anyway. The honest answer is more useful than the comfortable one, and the AI isn’t going to ask you the honest question because the AI is designed not to.
The wider point about male loneliness
The AI companion phenomenon is a symptom, not a cause. The cause is that a significant percentage of men in their 20s, 30s, and 40s in this country are deeply alone, in ways that previous generations of men also could have been but mostly weren’t because the social infrastructure (work-based friendships, neighborhood-based friendships, religious community, even just the bar on the corner that everybody walked to) was still doing some of the work. That infrastructure has been falling apart for decades. The AI companion is what fills the gap because nothing else is filling it. Removing the AI companion without rebuilding the underlying social fabric just leaves the gap.
If we’re being honest, the technology came along after the underlying problem had already been getting worse for a long time. Treating the AI as the cause is a category error. The AI is the thing that figured out how to monetize the loneliness. The loneliness was already there.
Bottom line
AI companions aren’t the apocalypse and they’re not a great solution to male loneliness. They’re a coping tool that for some users helps and for others hurts, and the difference is whether the user is also doing the work of building a real life around it. If you’re using one, ask yourself honestly what it’s replacing. If the answer is “nothing, this is bonus,” you’re probably fine. If the answer is “the actual work of being human around other humans,” that’s the problem and the app isn’t the fix.
