A user opens Rizz, an app built specifically to generate opening lines and replies for dating apps, pastes in a match's profile, gets three options, and sends one. On the other end, the recipient may be running the reply through a similar tool before answering. Neither necessarily knows the other is doing it.
A well-crafted opening line has always been a filter as much as an icebreaker, a small signal of effort or humor that a recipient uses to decide whether to invest further. When a model generates that line optimized purely for response rate, it can raise the number of conversations started while degrading the signal the line was supposed to carry. The recipient is often, in effect, evaluating the model's writing rather than the person's.
The people most exposed to this mismatch are the ones for whom the early message genuinely mattered as a preview of compatibility and personality. For them, an assisted opener trades a small amount of authentic signal for a real amount of throughput, a tradeoff that is not obviously worth it.
Once it becomes common knowledge that early messages might be assisted, recipients start rationally discounting them. Match Group has tested AI-assisted messaging features across parts of its portfolio, and Grindr has publicly discussed building an AI "wingman" concept; as these features spread, a witty opener increasingly reads as evidence of a good model rather than a good match. That discounting pushes genuine evaluation later, to a voice call or an in-person meeting.
What resists automation is the part that actually determines whether a relationship works: the willingness to disclose something slightly risky, in one's own words, and respond to another person's vulnerability with attention rather than a polished reply. No model can do that disclosure on a user's behalf without the exchange stopping being what it claims to be.
Rizz can get someone to reply. It cannot get two people to a first date that goes anywhere, because the part it skips, the risk of writing something true yourself, was never the friction the apps were trying to remove. It was the point.
