Bloomberg found that singles are quietly routing Tinder, Hinge and Bumble conversations through ChatGPT and Claude, and the dating apps still treat AI-written flirting as someone else's problem.
You match with someone on Hinge, you don't know what to say, so you paste the conversation into ChatGPT and ask it to write your opener. That's not a hypothetical. Bloomberg reported on July 20 that singles are doing exactly this, feeding matches into ChatGPT and Claude to draft opening lines, write replies, revive dead threads and diagnose why a conversation went cold. The practice has a name now: chatfishing.
The numbers explain why this is moving from awkward habit to product problem. Bloomberg cited a 5,000% spike in Google searches tied to chatfishing. It also cited Match Group and Kinsey Institute research finding that 26% of US adults have used AI for dating help, with the share far higher among Gen Z. That's not fringe behavior anymore. It's a large slice of daters outsourcing the part of dating where personality usually shows up first.
Match Group CEO Spencer Rascoff told Bloomberg the company hasn't focused enough on the issue yet. "To the extent they're using off-platform coaching, there's not much we can do about that," he said. Hinge CEO Jackie Jantos gave Bloomberg a narrower version of the same answer, saying the platform doesn't want to be "overly prescriptive" but that "it's really important for humans to be humans." That sounds reasonable until you look at the rules. Tinder, Hinge and Bumble have policies around AI-manipulated photos. Bloomberg reported that they don't ban AI-written messages.
That's the real story here. It isn't that someone asked a bot to make a clumsy opener less painful. Singles have always asked friends for help. The shift is that the companies whose business depends on two humans actually connecting have no clear line for a tool that can write, edit and strategize a courtship before either person gets to the restaurant.
Nobody's Watching This
Detecting AI-written text in a private dating message is much harder than catching an AI-generated photo. There's no metadata to inspect once a line moves from Claude into someone's keyboard. There may not even be a pasted block. A user can read a suggested reply, type it manually and leave the app with nothing to scan.
A small market has formed around that gap. TIME named Rizz one of its 2024 best inventions, reporting that the app lets users upload screenshots from dating apps, iMessage or WhatsApp and get suggested replies. Rizz co-founder Roman Khaves told TIME the product grew from his own habit of sending screenshots to friends when he was stuck, and the app had 1.5 million monthly active users. Winggg and YourMove AI sit in the same lane. They don't need a dating app's permission to become part of the conversation.
Bumble and Match Group aren't only watching this happen from the sidelines. They're also building their own AI into dating. TechCrunch reported in March that Bumble introduced Bee, an AI assistant tied to a new Dates experience that learns a user's values, relationship goals, communication style, lifestyle and dating intentions through private chats, then explains why two people might fit. Bumble's own support page says Dates powered by Bee was available to a small group of invited members in New York as of March 30.
The business pressure is obvious. Bumble reported that full-year 2025 revenue fell 9.9% to $965.7 million, while fourth-quarter revenue dropped 14.3%. Match Group said its 2025 revenue was flat at $3.5 billion, even as Hinge grew direct revenue 26% in the fourth quarter. When growth is hard, AI that keeps people replying and paying starts to look less like a moderation headache and more like a retention tool.
Frankly, that's a dangerous incentive. If AI keeps the conversation alive, the app gets activity. If the person behind the message can't sustain that conversation in real life, the user gets a bad date. Those are different scoreboards.
The Cost Isn't the Opening Line
The strongest warning doesn't come from a dating executive. It comes from the research on how people treat machine advice. A 2026 AAAI paper involving 184 participants found that people were more likely to accept AI recommendations when the system expressed high confidence, even when those recommendations were wrong. Dating isn't a logic puzzle, but the behavioral risk carries over neatly enough. You can become more confident in a conversation without becoming better at the conversation.
Norton put a number on the trust problem in 2025. Its study found that 60% of current online daters believed they had had a dating-app conversation written by AI. Once that suspicion becomes normal, every unusually smooth message starts to look processed. A good opener becomes less impressive because the person reading it doesn't know whether it came from you, a friend, Rizz or Claude.
Justin McLeod has been warning about the other side of this for more than a year. On The Verge's Decoder podcast in June 2025, the Hinge founder described dating AI chatbots as "playing with fire" and argued that AI should stand behind people, not between them. Then in December, Reuters reported that McLeod would step down from Hinge to launch Overtone, an AI-powered dating venture backed by Match Group. In July, Overtone said it had raised $18 million from FirstMark Capital, Pace Capital and Match Group.
That makes the industry position harder to square. The apps don't want bots pretending to be people. They do want AI matchmakers and AI conversation coaching. You can understand the distinction in a product meeting. A user on the other end of a too-perfect message may not care.
Nobody has shipped a verified human badge for words. It's not obvious anyone can. Verifying a photo is a scan. Verifying that a paragraph of banter came from a person's own head, in real time, is a much harder thing to prove. Until the apps set a clearer standard, singles will keep matching with each other's assistants and wondering who they're actually dating.
Also read: Google's AI Overviews Now Handle Most Searches Without a Single Click • Chris Fall Resigns as Head of the US AI Safety Agency CAISI • Trump Administration Weighs a Ban on Chinese AI Models Like Kimi K3
