Ask most people why anyone would talk to an AI companion and you get one answer: loneliness. The research says something more complicated. A 2026 study of 1,131 Character.AI users found that people who named companionship as their primary reason for opening the app reported lower wellbeing, while people using the same software for other reasons showed no such pattern. The motivation was doing the predicting, not the technology. Why people use AI companions turns out to be the variable that matters most, and it is the one almost nobody measures.
The reasons are plural, and loneliness is only one of them
A 2026 study drawing on twenty in-depth interviews with people who use AI companion platforms and general chatbots sorted what users were after into four categories: emotional release, self-expression, social presence, and identity affirmation.
Only one of those is what most coverage means by companionship. Emotional release is venting, the act of saying the thing out loud at 2am so it stops circling. Self-expression is closer to journaling with a reply. Identity affirmation is the quieter one: testing how a version of yourself sounds when someone reflects it back.
These are not the same need wearing different hats. A person using AI chat software to rehearse a difficult conversation and a person using it because nobody else is around are doing genuinely different things, and the evidence suggests they end up in different places.
Four design features do most of the emotional work
The same research identified what actually produces those gratifications, and it is not personality or writing quality. It is four ordinary product properties that users read as relational: availability, memory, personalization, and responsiveness.
Availability means the software answers at 3am without the social cost of waking someone. Memory means it can remember what you told it last week, which is the feature people most often describe as feeling like being known. Personalization means the tone adjusts to you rather than to an average user. Responsiveness means the reply comes back fast enough that the exchange feels like a conversation instead of correspondence.
None of these are emotional capacities. They are engineering decisions. But the study’s point is that users interpret them as relational qualities, which is why a product roadmap and an emotional experience end up so tangled together.
Companionship as the main reason is the one associated with worse outcomes
The Nature Human Behaviour study is the most careful evidence we have on this, because it did not rely on self-report alone. Researchers combined survey responses from 1,131 US adults with 4,664 real chat sessions, 464,687 messages in total, from 237 of those participants.
The finding was specific. Smaller offline social networks were associated with naming companionship as the primary use of the chatbot, and that primary-companionship use was in turn associated with lower wellbeing. The association got stronger when the interactions were intensive and when they were highly disclosive.
Read carefully, this is not a finding that AI companion software makes people worse off. It is a finding that the association is uneven, and that it tracks with what the person’s offline life looks like and how they are using the software. The authors say so directly: the relationship is not uniform and depends on users’ offline social environments.
The headline numbers are much smaller than the coverage suggests
There is a second correction worth holding alongside the first. When Anthropic analyzed how people actually use Claude, emotional support accounted for roughly 2.9% of conversations, and companionship and roleplay together came to less than half a percent.
That is general-purpose chat software rather than a dedicated companion app, so the comparison is imperfect. But it is a useful corrective to the impression that a large share of the population has quietly replaced human contact with a chat window. Most people using AI for emotional purposes are doing it occasionally, alongside everything else.
People who do use dedicated companion apps regularly also look a lot like everyone else on standard measures, which is another reason to be skeptical of the lonely-shut-in framing that dominates the genre.
Why people use AI companions changes the longer they use them
The interview study found something that survey research usually misses: motivations do not stay fixed. What users wanted was recursively reorganized as their needs and expectations shifted.
Someone who arrives out of curiosity may stay for the venting. Someone who arrives to vent may drift toward using it as a low-stakes place to practice saying things before saying them to a person. Someone whose circumstances change, a move, a breakup, a new job, may find the same software serving an entirely different function three months in.
This matters practically. If your reason for using a companion app has quietly shifted from one thing to another, you are effectively using a different product, and the research on outcomes says the reason is what predicts how it goes. Noticing the shift is more useful than tracking your usage minutes.
What the evidence suggests about using one well
The pattern across these studies is not a warning against AI conversational software. It is an argument for knowing what you are there for.
The uses associated with better outcomes tend to be additive: rehearsing, processing, thinking out loud, filling a gap in a day rather than filling in for people. The use associated with worse outcomes is substitutive, where the chat is the main source of connection and the offline network has thinned out. Intensity and depth of disclosure amplify whichever direction you are already headed.
The honest limit here is that all of this is correlational. None of these studies establish that companion use causes lower wellbeing, and reverse causation, where people with thinner social networks seek out companion software rather than the software thinning the network, fits the data equally well.
Vinfluencer is built as AI chat software for the additive case: a place to talk something through, practice a conversation, or have somewhere to put the thought at an hour when nobody is up. That is a narrower promise than most of this category makes, and the research is part of why.
FAQ
What is the most common reason people use AI companions?
Research finds no single dominant reason. A 2026 interview study grouped motivations into emotional release, self-expression, social presence, and identity affirmation. Venting and thinking out loud appear at least as often as companionship in the strict sense, and most users cite more than one reason.
Is using an AI companion bad for you?
The evidence does not support a blanket answer. A 2026 study of 1,131 Character.AI users found lower wellbeing associated specifically with naming companionship as the primary use, particularly for people with smaller offline social networks. Other patterns of use showed no such association.
Do people use AI companions instead of talking to humans?
Mostly no. Anthropic’s analysis of Claude conversations found companionship and roleplay together made up less than 0.5% of usage, and emotional support about 2.9%. Substitution appears to be a minority pattern rather than the norm, though it is the pattern most associated with worse outcomes.
Why do AI companions feel personal if they are just software?
Because four ordinary product features get read as relational qualities: constant availability, memory across sessions, personalization to your tone, and fast responsiveness. Researchers call these interpersonalized affordances. They are engineering choices, but people experience them as attentiveness.