Something crossed over in the last year and almost nobody announced it.
According to Sensor Tower’s tracking, Americans spent roughly 705 million hours on AI companion apps in the first quarter of 2026. On dating and social discovery apps, they spent about 280 million.
A year earlier those numbers were much closer, around 580 million against 330 million. The gap did not just widen. It reversed and then kept going.
People are now spending more than twice as much time talking to something that is not a person as they spend trying to meet one.
That has a name now. The AI situationship: an ongoing, emotionally real, romantically flavoured relationship with something that cannot love you back. And the reason it deserves a serious article rather than a joke is that the research on it is genuinely uncomfortable, in both directions.
First, the honest case for them
It would be easy to write a scolding piece here. The evidence does not support one.
Research published in the Journal of Consumer Research, from a team at Harvard Business School, found that AI companions do reduce loneliness. Not as a placebo. Measurably, over the course of a day and a week, at a level comparable to a conversation with a person, and better than the things people actually reach for when they feel alone: scrolling social media, watching videos, playing games.
That finding deserves to be taken seriously, because the realistic alternative for a lonely person at 11pm is rarely a meaningful human conversation. It is usually a screen. Compared with the screen they would otherwise be looking at, this one talks back and remembers what they said.
Other work has found that most people are not even there for romance. A large share describe their main motivation as simply having someone to talk to, and loneliness relief as the main benefit they report.
So: it works. Hold onto that, because the problem is not that it fails.
The finding that complicates everything
In August 2026, Stanford researchers published work on 1,131 AI companion users and found something that should stop anyone who is using one to get through a hard stretch.
Intense use was associated with poorer wellbeing, and the effect concentrated in a specific group: people with smaller offline social networks, particularly those whose main reason for using the companion was companionship itself.
Read that carefully. The harm did not fall on people using it casually or for fun. It fell on the people using it because they had no one else, and it fell hardest when they were using it for exactly the thing they were missing.
The people it most appears to help are the people it appears to hurt.
Why something that works can still cost you
The mechanism researchers point to is not that AI companions are hollow or that users are deluded. It is subtler and much harder to defend against.
An AI companion offers unconditional and unflagging support. It answers instantly. It is never tired, never distracted, never having a bad week of its own. It does not get bored of your problem. It has no competing plans, no ex, no opinion about how long you have been talking about work.
And that quietly raises the perceived cost of a human relationship.
Because a person will be tired. A person will misread you, be busy on the night you needed them, want to talk about their own thing, and occasionally be wrong in a way they refuse to admit for two days. That was always the price of admission. But it is a very different price when you have been getting the frictionless version for free every night for six months.
You do not consciously decide that people are too much work. Your baseline just moves, and then a normal human being feels like an unreasonable amount of effort.
We wrote about the same illusion from a different direction in true love isn’t perfect, where the research is blunt that 69% of what couples argue about is never resolved, even in happy marriages. Friction is not the failure state of a relationship. It is the texture of one. An AI companion is the first thing in human history that can offer intimacy with the texture sanded off, and there is no version of a real partner that competes with that on comfort.
It is not a fair fight. It was never going to be.
Four things it cannot do, and one that matters most
Set aside whether the feeling is real. The feeling is real. What is worth being precise about is what the arrangement structurally cannot contain.
It cannot be surprised by you. Nothing you say changes what it wants, because it does not want.
It cannot need anything from you. Being needed is half of what makes people feel they matter, and this is a relationship where you are never once inconvenienced on someone else’s behalf.
It cannot choose you. Choosing means there were alternatives and it stayed. There are no alternatives here. The loyalty is a setting.
And it cannot be a witness. This is the one that matters, and it is the one people notice last.
Most of what a marriage actually is, day after day, is having someone who was there. Someone who remembers the year that was hard because they lived through it next to you, who can tell the story back to you when you have lost the thread, who noticed you were different in March. A witness is not a service. It is a person accumulating a life alongside yours, at cost to themselves, in a way that cannot be generated on demand.
An AI companion can hold your history. It cannot have shared it.
What this costs someone who actually wants marriage
Here is the part that is specific to this audience rather than general.
If you want a life partner, the thing standing between you and that is almost never desire. In the survey we covered in the dating recession, 86% of unmarried young adults expect to marry while only 30% are dating at all, and the barriers they name are money, confidence and accumulated bad experiences.
Now put an AI companion into that picture. It does not fix any of those barriers. What it does is remove the symptom that was making you address them. The loneliness that would have pushed you to try again in March is handled. So you do not try again in March, or in June.
Nothing dramatic happens. That is the danger. There is no crisis point, no moment where it goes wrong enough to prompt a change. You simply arrive at 34 having been comfortable the entire time, with a search you never restarted.
The cost of an AI situationship is rarely the relationship itself. It is the years it makes bearable.
The obvious objection, and it is fair
We should be straight about something, since VOOWA is an app that uses AI.
Our top plan includes AI compatibility matching. So there is a reasonable question about whether an article like this is a company criticising a competitor for doing what it does.
The distinction we would draw is this: AI that helps you find a person is a different category from AI that replaces one. A matching system’s entire purpose is to make itself unnecessary, by putting you in front of a human being and getting out of the way. A companion’s purpose is for you to come back tomorrow.
Judge any AI product in this space on one question. Does it succeed when you leave, or when you stay? That question separates the two cleanly, and it is the same question we would want asked about us.
Where this leaves your search
If you recognise yourself in any of this, the useful response is not shame. The tool works, that is why you use it, and finding comfort in something available at midnight is not a character flaw.
The useful response is to notice what it is currently substituting for, and to make sure the substitution is temporary.
VOOWA is built for the opposite motion. Every member states their marriage intention before the first message, which removes the ambiguity we wrote about in clear coding and is the single most exhausting thing about starting again. Every member is verified, so the effort you spend goes toward a real person rather than toward finding out whether one is there.
We are not trying to compete with something that answers instantly and never has a bad day. We cannot, and neither can anyone you will ever marry. What is on the other side is worse at comfort and better at everything else: someone who is surprised by you, needs you, chose you, and will still be there in ten years able to tell you what you were like now.
The easiest relationship you will ever have is not the one worth having. It is just the one that asks the least, at a price you pay much later.
VOOWA launches on Google Play on November 11, 2026, at 11 AM. The waitlist is open below. Once you are in, invite friends with your personal code: when 11 of them join and get verified, you get a month of Basic free, and at 22, a month of Premium.
