Chatfishing is using AI — usually ChatGPT or a wingman app like Rizz — to write your dating-app messages, so you come across as wittier, more insightful, and more emotionally fluent than the real person behind the profile. It's the texting-era cousin of catfishing, with one key difference: the person is real. The photos are theirs, the name is theirs, the face that shows up to dinner is theirs. The personality you fell for over text is the part that was borrowed. As one psychologist put it, chatfishers usually aren't pretending to be someone else — they're trying to be themselves, but better than they can actually pull off.

The three most reliable tells, up front:

  • Uncanny polish. Casual texting is messy. Chatfished messages aren't — they're structured, grammatically immaculate, and just a little too composed for someone supposedly firing off a reply between meetings.
  • Warm but weightless. The replies sound attentive and emotionally generous, yet they mirror your words back without adding anything specific. Lots of "I love that you said that" and very little of their own concrete detail.
  • The charm has a delivery format. Move to a phone call or video, and the effortless wit disappears. The texter and the talker are two different people.

What Does Chatfishing Look Like?

  • The "perfect" opener. A first message so tailored to your profile it reads like it was written about you — because, in a sense, it was. Your bio went in, a flattering paragraph came out.
  • Reply latency that doesn't match the polish. A 40-minute gap followed by a flawless three-sentence response. Real charm is faster and sloppier than that.
  • Therapy-speak fluency. Effortless emotional vocabulary — "holding space," "I really hear you," "that must have felt invalidating" — deployed a little too smoothly for week one. (See also: therapy-speak used as a costume.)
  • Vagueness under follow-up. Ask them to expand on something specific they "said," and the detail evaporates. AI is great at sounding interested and bad at remembering what it pretended to care about.
  • A voice that resets. The texting style shifts noticeably from day to day, because the prompt — or the person copy-pasting — shifts too.

Where Does the Term Come From?

"Chatfishing" surfaced in 2025 as generative AI quietly moved into the messaging layer of dating. Writing in Scientific American, Deni Ellis Béchard framed it as a modern Turing test playing out one match at a time, and cited the scale of the problem: a Match survey found a sharp year-over-year jump in singles using AI to enhance their dating, and a Norton report found that roughly 60% of dating-app users believed they'd already encountered AI-written conversations. The detection problem is real — Béchard notes that in controlled studies people guessed AI versus human text barely better than a coin flip.

Psychologist Bruce Y. Lee, writing for Psychology Today, is blunter about the stakes: there's no reliable software to catch it yet, the usual "tells" are nowhere near 100% accurate, and the willingness to outsource your charm is itself a data point about trustworthiness. The deeper problem isn't the grammar. It's sequencing, the same trap as kittenfishing: the first thing a date discovers about a chatfisher is that the connection they felt was rented.

In Practice

Three weeks of the best texting of her life. He was funny in that precise, surprising way — callbacks to things she'd mentioned, a knack for naming exactly what she meant before she finished saying it. She'd reread the messages. Then they got on the phone, and the man on the other end was perfectly nice and completely flat: long pauses, generic answers, none of the spark. On the date it was worse — she'd reference a joke from their texts and he'd half-laugh like he didn't quite remember it, because he didn't write it. The chemistry hadn't curdled. It had never been his. The painful part wasn't that he wasn't witty. It was that she'd spent three weeks falling for a draft.

What to Do About It

Move off text early. The phone-or-video call in week one is the single best diagnostic. A real personality survives the format change; a borrowed one doesn't.

Ask a question only the human can answer. Reference something specific they said three messages back and ask them to go deeper. AI is fluent in the moment and amnesiac about it afterward — the person should remember their own life.

Read the live version, not the transcript. Once you've met, weight the in-person human over the texter. If you keep reaching for the screenshots to find the spark, the spark may not be in the room.

Decide what you actually mind. A grammar fix isn't deception. An outsourced personality is. Get clear on which one happened before you write someone off — and notice the red flags that were in the texting all along.