A quarter of American singles use AI in their dating lives, half of Gen Z does, and the apps say they cannot stop it. When both people on a date have been coached by the same model, the thing that meets for coffee is a conversation between two drafts. The person who shows up is the one neither side has met.
Match and the Kinsey Institute surveyed 5,001 American singles for the 2025 edition of Singles in America and found that 26 percent had used AI in their dating lives, a 333 percent rise in a year. Among Gen Z the figure was nearly half. The uses were the ones you would guess: building a better profile, crafting a stronger opener, screening a match for compatibility. Match’s director of relationship science called it an edge rather than a replacement. “AI isn’t replacing intimacy,” Amanda Gesselman said, “it’s giving singles an edge.”
A year on, the practice has a name, chatfishing, and a different tone. The Next Web reported in mid-2026 on singles pasting whole dating-app conversations into ChatGPT or Claude and asking for the reply, the analysis of what went wrong, the next line. A woman in Virginia noticed her Hinge match’s phrasing was off; he admitted the messages were Claude’s. A woman in San Diego described a phone call in which the man appeared to be reading a script a chatbot had generated from her profile. None of the major apps prohibit it. Match’s chief executive told Bloomberg that off-platform coaching is beyond the company’s reach. Hinge’s said it did not want to be “overly prescriptive.” Bumble does not screen for it. No app detects it, and the detection tools that exist are unreliable.
So one in four, and rising, and nobody at the door.
The edge that cancels out
The arithmetic of an edge is that it only works if the other person does not have it. When one in four singles is drafting with a model, a large share of conversations on the apps now involve two people being coached by systems trained on the same corpus toward the same approved register. Every AI sounds the same for a reason that is arithmetic rather than aesthetic: tuned on human approval, models converge on the response nobody dislikes, which is the average. Two singles each using that average to impress the other are sending the same person’s messages back and forth and calling it chemistry.
The Virginia woman noticed because the phrasing was too smooth. That is the tell, and it is the whole problem compressed. What made her match a specific person, the slightly wrong word, the joke that only he would make, the thing he noticed in her profile that nobody else did, was exactly what the model removed. He was more fluent and less there.
What the coffee is for
The anecdote that matters most is the smallest. A woman whose date admitted to using ChatGPT for his messages said the in-person conversation was awkward. Of course it was. The messages had been auditioning a person who did not exist, and the person who arrived had never rehearsed the role.
Dating apps have always had this gap. The profile is a performance and the date is the test. What AI changes is the distance between them. A person can now sustain a two-week text conversation at a level of wit, attentiveness, and emotional fluency that they cannot produce for ten minutes across a table, and the other person has been forming an attachment to that level. The coffee used to be where you found out whether the chemistry was real. Now it is where you find out whether the person you have been talking to exists.
This is the same failure the companion category sells on purpose, arriving from the other direction. A companion app gives you a counterpart who is always fluent and never there. Chatfishing gives you a real person wearing that counterpart’s voice, until the moment they have to speak for themselves. In both cases the thing you connected with was an average with no one behind it, and in both cases the discovery comes late and costs something.
Why the apps cannot stop it
Match’s chief executive is right that the company cannot police what happens in another app on the same phone. But the deeper reason the platforms are not trying is that chatfishing is good for their metrics. A smoother conversation is a longer one. A more confident opener gets more replies. An AI that keeps a hesitant user texting is doing the retention work the app would otherwise have to do itself. The platforms have no incentive to detect the thing that makes their engagement numbers look better, and the honest statement from Hinge, that it does not want to be prescriptive, is the sound of a business noticing that.
Meanwhile the one attitude figure in the research cuts the other way. A 2026 Match survey of younger singles found that two in five would refuse to date someone who uses an AI companion app, rising to 51 percent among women aged 18 to 24. The same people who draft their messages with a model do not want to date someone in a relationship with one. They have drawn the line at the counterpart, and not yet noticed that they have let it write their half of the conversation.
The person who shows up
There is a simple test for whether a tool helps or hurts in dating, and it is whether it moves the person closer to the table or further from it. Filtering matches by compatibility moves them closer; it gets two real people in front of each other faster. Writing the messages moves them further, because it makes the two weeks before the table a conversation with someone else, and then charges the table with the job of revealing it.
The deeper point sits under both. People are using these tools because thinking out loud needs someone, and a model is the someone that is available at the moment you are staring at a blank reply box. What it cannot do is be the one who meets her. The counterpart worth having in that moment is one with a view of its own about what you are trying to say, one that would say the joke is not yours and the line is not true, and then get out of the way so that the person who shows up for coffee is the one who wrote the messages. Anything else is a very polite introduction to a stranger.
One in four singles has decided the risk is worth it. Nearly half of a generation has. They are, collectively, meeting an average, and the average cannot show up.
Sources: Match Group and The Kinsey Institute, Singles in America, 14th annual study (June 10, 2025; 5,001 US singles, Dynata; 26 percent, 333 percent, Gen Z near half; Amanda Gesselman quote). The Next Web, “AI is writing your date’s messages. The dating apps say they cannot stop it,” Ana Maria Constantin (July 20, 2026; chatfishing anecdotes; Spencer Rascoff via Bloomberg; Jackie Jantos; Bumble; detection tools). Axios (July 14, 2026; Match Group survey of about 1,000 singles aged 18 to 39: two in five would refuse to date an AI-companion user, 51 percent among women 18 to 24). Vice and Global Dating Insights (2026, chatfishing coverage).
