How the face attractiveness test works

Your photo stays on your device. Your browser finds your face, and you get a score from 60 to 99, a band and the three features that stand out most.

How the score is made

The test uses an AI model that learned from many faces people rated. It compares your photo with a large set of rated portraits and places it on a scale from 60 to 99, where 80 is the middle.

On faces it never saw while learning, its scores line up with people’s average ratings at a correlation of 0.84, where 1 would be a perfect match. Those ratings had the raters’ preference for lighter skin taken out first, so the test does not learn it. On a harder, more varied set of faces the agreement is lower, about 0.6, and it is similar across the groups of faces we checked.

If your photo lands in the upper half, you also see a Top X% figure. Top 20% means it scored higher than 80% of the portraits it is compared with, so it is always good news.

Everything runs in your browser. The photo is never uploaded, and the result appears in a few seconds. Privacy, explained ›

What the four bands mean

Every score falls in one of four bands. The scale starts at 60 on purpose: a photo is not a verdict on a person, so there is no failing grade. These bands belong to this test's own 60–99 scale, so a score here is not directly comparable with another site's 1–10 rating.

BandScore
Natural Charm60–69
Glowing70–79
Radiant80–89
Dazzling90–99

Highlights and readings

Your browser measures a few proportions of your face, such as eye size, eye spacing and jaw width, and compares each with the typical range across a broad mix of people. The three that stand out most become your highlights, and they only ever name what works on your face.

A “than most” highlight compares you with a broad mix of people, with every background, men and women counted equally. Not every face gets one, and a balanced highlight is just as good. Some groups get a “than most” highlight more often than others, which is a known limitation of comparing everyone with one mix.

Five readings describe your proportions in neutral words, like “Eye spacing: balanced”. Highlights and readings are descriptions, not scores. They do not add up to your number, and there is no right answer.

Skin is never measured. In a photo, skin mostly shows the light, the camera and any filter, and judging it is where bias about skin tone creeps in. The highlights were checked on ordinary phone selfies as well as studio portraits, and readings that did not hold up on selfies were left out.

What moves your score

The test scores the photo, not the person. Light, head angle, expression and distance from the camera all change how a face reads, so three photos of you can get three different scores. In checks on the current model, photos turned to the side scored about the same as front-facing ones. When your head is turned well to the side, the test tells you and skips the readings that need a straight-on view.

Age plays a part too: older faces tend to score a little lower, because the people who rated faces tended to prefer younger ones, and most faces the test learned from belong to people aged 20 to 40.

For the steadiest score, face the camera in soft light from the front, at eye level, about an arm’s length away.

Limits and fairness

The same photo always gets the same score. Everyone gets the same scale and the same wording, and you are never asked your gender.

To test skin tone on its own, we took 533 face photos, made only the skin lighter or darker by 12 and 20 steps on a 100-step lightness scale, in two different ways, and scored each photo again. On average the score moved by less than 0.1 points, and for 9 in 10 photos by about 1.2 points or less, but a few single photos moved by as much as 4.5 points. Between ethnic groups, the highest and lowest average scores differ by about a fifth of the usual spread between people in the same group for women, and by about a third for men. Within two groups the score still leans slightly with skin tone, in opposite directions: darker skin scores a little lower among Indian faces, and lighter skin a little lower among East Asian faces. These checks used everyday photos of people, so they are strongest for photos like those.

A score describes how one photo reads to a large group of people. How you move, sound and laugh is bigger than any photo. The score is not a measure of your worth, and it is not medical or psychological advice.

Ready to take the attractiveness test?

Take the test

Free, and your photo stays on your device.