How it works
KNOWN asks you a set of questions and turns your answers into a written portrait of the kind of person you are looking for. This page explains what actually happens in between, including where a language model is used and where one is not.
The three parts
Talk. Tell us what you’re looking for.
Reflect. Tell us what you’ve wanted — and what hasn’t worked.
Discover. Meet the person your answers describe.
The questions are written by hand
Every question you are asked was written by a person and is the same for everyone. Nothing decides what to ask you next based on what you just said, and no model writes a question.
That is a deliberate choice rather than a limitation we are admitting to. It means the questionnaire is the same length for everyone, it can be read and improved by hand, and there is no chance of it drifting somewhere strange because of one unusual answer.
Where the AI is used
KNOWN’s AI analyses your answers to reveal the patterns, priorities and relationship dynamics that matter most to you.
Concretely, a language model is used in four places: to read each answer and file it into a structured profile, to write the short reading you see after the opening questions, to propose the three names you choose from, and to write the portrait and the report.
It is never used to decide what to ask you, to score you, or to compare you with anyone else.
What the portrait is
It is built only from what you told us. Nothing about your browsing, your social accounts or anyone else’s answers goes into it.
It describes an ideal — a coherent picture of the kind of person and the kind of relationship your answers point to. It is not a prediction about a particular man, and the report says so in its own words.
What KNOWN is not
It is not a dating app and not a matching service. There is no database of real people here, nobody is introduced to anybody, and no part of this looks for a person on your behalf.
It is not therapy, counselling or advice from a qualified professional, and it is not a test with a right answer.
Whether it can be wrong
Yes. A language model can write something confident and inaccurate, and a portrait built from a short questionnaire is a reading rather than a fact about you. If a line does not fit, it does not fit.
The portrait is written to be specific enough to be disagreed with, which is the only kind worth reading — but that also means some of it will occasionally miss.
Where your answers go
Your answers are stored so you can come back to your portrait, and they are sent to the model provider only to produce it. They are never used to train anyone’s model. The privacy page lists every company involved and what each one does.