Face apps love to hand out verdicts: "your face is 91% symmetrical", "you match the golden ratio at 87%". These numbers feel scientific and mostly aren't. Here's what the research does and doesn't support, and why FaceTest deliberately measures something else.
Myth: perfect symmetry is what makes a face attractive
Symmetry does correlate with how attractive people rate a face, but the effect is small, and it's tangled up with other things: symmetric faces tend to also be closer to average and to look healthier. Every real face is asymmetric; one eye sits a little higher, one side of the mouth rises more when smiling. And the composite experiment tells the real story: take a face, mirror the left half to make a perfectly symmetric "left-left" face and the right half to make a "right-right" face, and both usually look slightly wrong, even uncanny, compared with the original. Our brains expect asymmetry. A little of it is part of what makes a face look alive.
Myth: the golden ratio face
The idea that beautiful faces follow the golden ratio (about 1.618) in their proportions is popular in cosmetics marketing and in "the most beautiful face according to science" articles. It doesn't hold up. The ratios chosen are picked after the fact from the many possible measurements on a face, different proponents pick different ones, and controlled studies find no special preference for 1.618 over nearby values. The "golden mask" overlays that circulate online fit well-known faces about as well as any other symmetric template would. It's a nice number; it's not a law of faces.
Myth: attractive means distinctive
The opposite is closer to true. In a classic 1990 experiment, Judith Langlois and Lori Roggman digitally averaged many faces together and found that the composites were rated more attractive than most of the individual faces that went into them: the averageness effect. Faces near the population average look familiar and easy to process, and we read that ease as attractiveness. The very most attractive faces are not perfectly average, though; they tend to be average with a few features exaggerated in a particular direction. So both things are true at once: typical is attractive, and a bit of well-placed distinctiveness is more so.
Myth: symmetry apps tell you something
An app that says your face is 91% symmetrical is comparing your left and right halves in some way it doesn't specify, from a single photo where a slight head turn or an uneven light source can shift the number by several points. Without a baseline (what do other faces score under the same method?) the number has no meaning. It's a random-ish figure with a confident decimal.
Myth: the AI is judging your looks
This one matters for us. A look-alike test measures distance in feature space between your photo and a set of character stills: it asks "which of these is shaped most like you", never "how good do you look". A symmetric face and an asymmetric face both get matched to whichever character shares their shape language. There is no hidden beauty score. If a result seems flattering or unflattering, that's your own reaction to the character, not a verdict from the model. How the matching works shows what the engine actually computes.
What's real
Faces are perceived holistically, typical faces are processed easily, and small asymmetries are normal and expected. Those are the findings that survive. The rest, the magic ratios and precision percentages, are marketing. Enjoy the face tests as what they are: a measurement of resemblance, with the judgment left entirely to you.