In the United States, 25% of young men say they feel lonely, against 17% of all other adults, Gallup reported in 2025 from its Meta-Gallup social connections data. The US is one of only three wealthy countries where young men are lonelier than everyone else. Companion chat software, meanwhile, has a user base that leans male. Put those two facts side by side and a tidy story forms: here is a problem, and here is the product arriving to meet it. The overlap is real. The fit is looser than it looks, and the difference is worth stating plainly.
Young men are the lonelier group, and the US is unusual in that
The Gallup figure is the one worth anchoring on, because it is a survey measure of felt loneliness rather than a proxy like time spent alone. Globally, young adults aged 19 to 29 report the highest loneliness of any age band, at 27%, while adults over 65 report the lowest. What makes the American picture distinctive is the gender split inside that young cohort. In most countries young men and young women report similar rates. In the US, Iceland, and Denmark, young men report more.
That is an eight point gap in the US, which is large for this kind of measure but not a crisis of a different order from everyone else’s. Any discussion of AI companions and male loneliness that opens with catastrophe is already off by a degree. The honest version is narrower and more useful: a specific group reports a specific shortfall, and it has grown.
What men report losing is close friends, not conversation
The most striking data here is not about loneliness at all. It is about headcount. The Survey Center on American Life found that in 1990, 55% of American men said they had at least six close friends. By 2021 that figure was 27%. The share of men reporting no close friends at all went from 3% to 15%, a fivefold rise. Women saw a decline too, from 41% to 24%, but a shallower one.
Read that carefully, because it constrains what any software can do. The thing that shrank is a roster of people who know you. It is not a shortage of talk, or of responsiveness, or of someone to describe your day to. Those are the goods a conversational companion app supplies well. A count of close friends is a different quantity, and no amount of conversation quality moves it directly.
This is the crux of why AI companions and male loneliness do not line up neatly. The apps are strong on the symptom that is easiest to feel at 11pm on a Tuesday. The measured deficit sits somewhere else.
Companion software is genuinely good at making someone feel heard
None of that makes the software useless, and the research on this point is better than the skeptical coverage suggests. In the Journal of Consumer Research, De Freitas and colleagues published a multi-study paper in 2026 finding that interacting with an AI companion reduced momentary loneliness on par with talking to another person, and considerably more than passive activities like watching videos.
The mechanism they identified is the part that matters. Across their studies, the strongest predictor of loneliness reduction was not how capable or fluent the model was. It was whether users felt heard. That is a modest, specific finding, and it is also exactly the good that a well-designed AI conversational companion application can deliver on demand at 2am, which no friend can.
Note the word momentary. The same paper is careful about it. The effect was measured after use, repeatedly, over a week. It is relief, reliably produced. Relief is worth something. It is not the same as a social circle.
The evidence gets more complicated among heavy users
Correlational work on students consistently finds that heavier use of generative AI for companionship travels with higher loneliness, and in US college samples the association holds after adjusting for depression, anxiety, and suicidality. That is a real finding and it is easy to misread in both directions.
It does not show that companion apps cause loneliness. Lonelier people plausibly seek out companionship software, which would produce the same pattern with the arrow running the other way. Cross-sectional data cannot separate the two. Anyone citing this research to prove harm is overclaiming, and so is anyone waving it away.
What it does justify is caution about volume. The pattern that shows up across this literature is that light and moderate use looks benign or mildly helpful, while the heaviest users are the ones reporting the worst outcomes. Whether that is selection or effect, it is a reasonable line to watch in your own usage.
The gap that software structurally cannot close
There is a plainer way to put the limit. A companion app is entirely on your side of the ledger. It listens when you want, remembers what you tell it, and never needs anything from you. That asymmetry is the product. It is also the reason the experience stops short of friendship, a question worth examining on its own terms if you have wondered whether an AI can be your friend.
Close friendship, the thing the friendship recession data actually tracks, runs on reciprocity. Someone notices you went quiet. Someone needs a favor and asks you specifically. You matter to a person who is not you. Nothing in a companion app generates that, because there is no one on the other side with stakes. For men who grew up with fewer scripts for asking directly for connection, this can be the most uncomfortable thing to hear and the most important.
Where companion software fits without pretending otherwise
The useful framing is rehearsal rather than replacement. Conversation with software is low-stakes, unlimited, and free of the cost of being awkward in front of someone whose opinion you care about. Research on AI companions for autistic adults points the same way: the value shows up when the software is used to practice and to decompress, and it thins out when it becomes the whole social diet.
Practically, that means a few things. Use it to get words around something before you say it to a person. Notice if a week goes by where the app was the only place you said anything real. Treat the relief it produces as relief, not as progress on the friend count. And if the count is the actual problem, the interventions with evidence behind them are unglamorous: repetition, shared activity, showing up to the same thing weekly until familiarity does its work.
Vinfluencer is built as AI chat software for exactly the first job, a place to talk and be heard without performance. It is not built to be the last conversation you have all week, and we would rather say so.
Frequently Asked Questions
Are most AI companion app users men?
Usage does skew male across major companion platforms, though estimates vary widely by app and none of the large platforms publish audited demographics. Reported splits have also narrowed over time as adoption broadened. Treat any precise percentage you see quoted with suspicion, because most trace back to company statements rather than independent measurement.
Does using an AI companion make loneliness worse?
The research does not support that as a general claim. Controlled studies find momentary loneliness drops after use. Survey studies find heavy users report higher loneliness, but cannot establish which came first. The reasonable reading is that moderate use appears benign and heavy substitution for human contact is the pattern worth avoiding.
Why is male loneliness discussed more than female loneliness?
Because the decline has been steeper. Both men and women report fewer close friends than in 1990, but men fell from 55% with six or more close friends to 27%, against 41% to 24% for women. In the US, young men also report higher felt loneliness than other adults, which is not true in most countries.
Can a companion app help someone build real friendships?
Indirectly at best, and only if it is used that way. It can lower the cost of practicing conversation and help someone articulate what they want to say. It cannot introduce anyone, and it cannot create the mutual obligation that close friendship is made of. If loneliness is severe or persistent, a therapist or a doctor is the better next step.