The process
How my type test works
Choose the adult faces that catch your eye. Your result summarizes recurring visible features in those choices, with avatars and counts you can inspect.
Choose Women or Men
Confirm that you are 18 or older, then choose an avatar pool. Every test portrait depicts an AI-generated virtual adult. You do not upload a photo or create an account.
The pool stays fixed during a test. To try the other pool, start a new test.
Work through 20 base crowds
Each crowd shows 12 avatars. Click every face you like and a new unseen avatar fills that position. Choosing none is valid when no one stands out.
Finish all 20 crowds to see a result. You can undo the most recent pick before leaving the current crowd, and continue in small blocks after the result when enough unseen faces remain.
- Avatars do not repeat within a test.
- You cannot return after leaving a crowd.
- There is no timer or required number of picks.
Read the evidence, then save
Your completed result shows picked portraits, unseen predicted matches, contrasting faces, recurring features, and support counts. A named type is used only when the choices support it; a mixed result is valid too.
Download a result card, copy a text summary, or use your device's share sheet when supported. A local result address opens only in the same browser; it is not a public sharing link.
Resume in this browser
Progress and results are saved locally in this browser. Refreshing lets you resume while that data remains available. Private browsing, clearing site data, or changing devices can remove access.
To remove local progress, results, and the remembered age confirmation, clear this site's storage in your browser settings. Keep an exported card separately if you want a copy.
Later crowds respond to earlier choices
The portrait order is deterministic for the same starting seed and action history, but it is not a fixed slideshow. Earlier likes change which visual directions receive closer comparison later. The test can place likely matches beside useful alternatives without repeating a face.
Adaptation does not mean the system knows your taste after the first click. Every portrait combines several visible features, so the session continues gathering both liked and unliked exposure before it summarizes a pattern.
Confidence reflects the evidence available
The engine compares liked and unselected faces within the same crowd. Every fifth crowd is held out of training and used as a small validation check. A result may remain provisional or mixed when the session contains too little consistent evidence.
These checks make the method inspectable, but they do not create scientific accuracy. The output remains a session-based description of visible choices, and the pictures are provided so you can disagree with the label.
Questions, answered
Is the basic result free?
Yes. The test and basic result need no payment or login. The optional Deep Report is a separate one-time purchase after you see the free result.
How long does it take?
That depends on your pace. The base task is 20 crowds with no required number of picks; no completion time is promised.