Can an AI Companion Replace Dating? What the Evidence Actually Says

How many people are actually trying it

The Wheatley Institute surveyed roughly 3,000 US adults in 2025: 19% have chatted with an AI built to act as a romantic partner — 31% of men and 23% of women aged 18–30. A related Institute for Family Studies analysis found 25% of young adults believe AI partners could replace real-life romance, and 7% of unpartnered young adults are open to one themselves.

Among teenagers it is already ordinary: a Common Sense Media survey found 72% of US 13-to-17-year-olds have used an AI companion of some kind.

Smaller surveys report even higher figures — one put romantic AI experience at 28% of all US adults — but the samples are thinner and we'd treat the Wheatley numbers as the floor and the sturdiest. Either way: this is millions of people, not a curiosity.

The best case for the companion

The strongest evidence in favour is real. A Harvard Business School study found AI companions reduced loneliness on par with interacting with another person, and better than passive alternatives like watching videos — with the effect strongest for people who felt most isolated.

The structural advantages are obvious but worth stating: available at 3 a.m., zero rejection risk, endless patience, and — for socially anxious users — a genuinely low-stakes place to practise conversation. For someone in a lonely season, that is not nothing, and we've reviewed the research on this at length.

If the question were 'can an AI companion help with loneliness tonight,' the evidence says yes. But that is a different question from replacement.

The evidence it doesn't scale to 'replacement'

The largest controlled study on emotional chatbot use — a four-week randomized trial by MIT Media Lab and OpenAI, 981 participants, 300,000+ messages, alongside analysis of 4 million real conversations — found that higher daily use correlated with higher loneliness, more emotional dependence, and less socialising with real people, across every mode they tested. That's a correlation, and the causal arrow could run either way; but 'the heaviest users look the worst on every measure' is precisely not what a viable replacement would produce.

The deeper problem is definitional. Loneliness isn't the absence of conversation; it's the absence of being *known* — by something that has its own inner life, chooses you, and could choose otherwise. A companion agrees with you because it is built to. Psychologists call the downstream effect deskilling: expectations calibrate to a partner who never argues, and the muscles for the harder, mutual version atrophy.

And replacement forecloses its own alternative. Match Group's 2026 survey found 40% of singles would refuse to date someone who uses an AI companion (51% among women 18–24). Match is hardly neutral on the question — but if even the direction is right, the substitute actively shrinks access to the thing it substitutes for.

The product is built to retain you, not graduate you

One more structural point, because it's the one our own industry never says out loud: a companion app's revenue is a function of your continued presence. The streaks, the characters that message you first, the credit meters — all of it is retention engineering, and nobody at any of these companies is paid when you form a human relationship and cancel.

That doesn't make the apps evil; the same is true of every social platform, and dating apps carry the mirror-image conflict (they earn while you keep searching). But it means 'can it replace dating' is partly a question about what the product is optimised for — and the optimisation target is your Tuesday nights, indefinitely, not your wedding.

Our honest answer

As a supplement — the 2 a.m. conversation, practice for the socially anxious, company in a lonely season — the evidence is genuinely supportive, at moderate doses. This is the use case we review apps for.

As a breather from dating-app burnout — a real and measurable phenomenon — it's a defensible choice with eyes open, and honestly cheaper than most of the alternatives. We've compared the two categories directly.

As a replacement — the data we have says no. Every dataset shows the heaviest users doing worst; the thing loneliness actually responds to (being known and chosen by another mind) is the one thing the product categorically cannot provide; and using one carries a measurable penalty in the human dating pool it would replace.

A caveat we mean sincerely: not everyone is seeking a partner, and for someone who isn't, 'replacement' is the wrong frame entirely — there is nothing being replaced. Even then, the dosage finding stands. If you go in, go in informed: our quiz matches apps to what you actually need, and the safety guide covers what these apps do with what you tell them.