What a modeled conversation can—and cannot—predict.
By Find Your Person · Published
Product-mechanics articles are checked against the current implementation and public policies.
Before a potential pair becomes an introduction, Find Your Person models a conversation between representations built from the information each member has provided.
The simulation follows several conversation phases and produces an evaluation that includes an overall score and an estimated second-date likelihood. That evaluation must clear a gate before the pair can move forward. It also contributes to the broader weighted match score.
It is one signal among several.
The simulation currently carries 28% of the baseline weighted score. Photo-preference fit, reciprocal discovery preferences, interview-derived traits, engagement, and confidence make up the rest. Both members’ directional scores must pass. The system is deliberately broader than a single AI-generated conversation.
It does not simulate two real people in full.
A model works from profile representations, not complete human beings. It cannot reproduce body language, timing, physical presence, attraction in context, or the countless surprises of an actual conversation. A high result means the available inputs look promising under the model. It does not mean a date will go well.
Why use it at all?
Written preferences can say whether two people want the same broad future. A modeled exchange can add a different kind of signal about topic overlap, pacing, and conversational fit. Used carefully, it can help rank potential introductions and generate a clearer explanation for why one is being shown. The member still decides whether to continue.