Why Do People Use AI Girlfriends? What the Data Actually Shows

Loneliness is real, but it's not the whole story

The 2023 US Surgeon General advisory on loneliness and isolation reported that around half of American adults experience measurable loneliness, with young adults among the worst affected. That statistic is the honest backdrop for this entire category. It doesn't mean every AI companion user is lonely in a clinical sense, but it does mean the gap these apps are stepping into is large, well documented, and predates the apps themselves by decades.

Treating loneliness as the single explanation flattens a much more varied set of actual users, in the same way that assuming everyone who reads romance novels is starved of affection would be a lazy and mostly wrong generalisation.

Low-stakes social practice

A meaningful share of users describe companion conversations as practice: rehearsing how to say something difficult, working through social anxiety in an environment with no risk of embarrassment, or simply getting comfortable with the rhythm of conversation again after a long stretch without much of it. This is closer to a low-cost rehearsal space than a relationship substitute, and it's one of the more defensible uses of the category precisely because the goal is usually to take the practised skill somewhere else.

Company during a specific life transition

Divorce, bereavement, relocation to a new city, a period of shift work that cuts someone off from friends' normal hours: these are the situations that come up repeatedly in reporting and user interviews as the actual trigger for trying an AI companion app, more often than a generalised, permanent loneliness. The pattern tends to be bounded rather than indefinite: usage rises sharply around the transition and settles once the person's social circle has rebuilt itself.

Plain curiosity and novelty

Not every user has a deeper reason. A large and often under-reported share are simply curious about what a modern language model can actually do in an open-ended conversation, in the same way people tried early chatbots like ELIZA or, more recently, general-purpose assistants, out of interest rather than need. Curiosity users tend to churn quickly, but they're a real and sizeable part of who signs up in the first place.

Fantasy and roleplay as a genre

The romance fiction market is worth well over a billion dollars a year and nobody frames every reader as lonely. AI companion roleplay sits in a similar cultural space for a meaningful share of its users: an interactive form of the same genre, enjoyed on its own terms rather than as compensation for something missing. Dismissing this motivation as sad rather than simply a preference says more about the stigma around the category than about the people using it.

What actually shifts once someone starts using one regularly

The more interesting research question isn't why people start, it's what changes once they keep going. Reporting on long-term users describes two roughly distinct patterns. In the first, the app functions as a supplement: it fills specific quiet hours, and the person's other relationships stay stable or even improve, because low-stakes practice on the app translates into more confidence offline. In the second, less common but the one clinicians and researchers flag as worth watching, the app starts crowding out other contact rather than sitting alongside it.

Nobody has a reliable way to predict in advance which pattern a given new user will fall into. What the research does suggest is that the difference has less to do with the initial motivation, loneliness versus curiosity versus practice, and more to do with whether the person keeps investing in relationships outside the app at the same time. That's a genuinely useful distinction, because it means the honest advice isn't 'don't use one if you're lonely', it's 'notice if it becomes the only place you're getting this from'.

Why the stigma around this doesn't match the data

A lot of the discomfort around AI companion apps assumes a single, embarrassing motivation, usually some version of 'they can't get a real relationship'. The actual spread of reasons above doesn't support that as a general claim. Someone rehearsing a difficult conversation with a parent, someone getting through the first lonely month after a house move, someone who just enjoys a well-written romantic scene: none of those map neatly onto the stereotype, and treating them as though they do mostly just discourages people from being honest about a genuinely varied set of reasons for trying something new.

It's worth noticing, too, how quickly the same conversation happened with earlier technologies. Online dating carried a near-identical stigma twenty years ago, framed as a last resort for people who'd failed at meeting someone the 'normal' way, right up until it became the most common way couples actually met. Video calling was once considered a poor substitute for seeing family in person, until distance and cost made it simply the sensible option. AI companionship is at roughly the stage online dating was in the early 2000s: genuinely useful for a real subset of people, widely used, and still carrying a reputation that lags a decade or two behind the actual behaviour.

What this means if you're using one, or thinking about it

None of the motivations above are a red flag on their own. Loneliness, curiosity, practice, a rough patch, plain enjoyment of the genre: all of these are ordinary, human reasons to try something, and the research doesn't support treating any of them as inherently unhealthy. The thing worth actually tracking, as the section above on long-term use covers, is whether the app is sitting alongside a life that's still being built with other people, or quietly becoming the only place that life is happening. That's a much narrower, more useful question than 'why do people do this', and it's one only the person doing it can really answer for themselves.