Phenotype

With that established, what it actually does.

Your face is turned into a numerical descriptor — a compact summary of its geometry — and that descriptor is compared against reference descriptors for the types catalogued in the humanphenotypes.net typology. That comparison is then checked against traits read from your own scan: a type whose recorded hair, skin or eye colour, nose breadth, profile or head shape plainly contradicts yours is moved down the list. You get back a ranked list of the closest types.

It is a rough estimate, and it is not part of your analysis: it never feeds your score, your pillars or your protocol.

The background you gave at the start of your scan — the groups you picked and any specific places — is used as a prior. Types centred where you said you are from rank higher; types found there only secondarily get part of that; types not found there at all lose a little ground. If you picked more than one background, each one counts.

It is a prior, not a gate: no type is ever excluded because of your answer, and a face that clearly resembles a type from somewhere else still ranks it first. In our tests on photos of people with a known background, answering lifted the rate at which the top type came from the right region from about 3 in 4 to 9 in 10 — and when the answer given was wrong, the face still won out about half the time.

If you did not answer, or picked only mixed heritage, the ranking runs on your face and measured traits alone.

Because it is interesting, and because it is honest about being only that.

It changes nothing. It contributes no points, gates nothing, and does not appear in any other result. If you find it useful as a way of describing your face, good. If you find it uninteresting, it costs you nothing to ignore — which is roughly the right level of weight to give it.