Facial symmetry is how closely the left and right halves of a face match each other when one is mirrored onto the other. That is the whole definition. A symmetrical face is one where the eyes sit at the same height, the two sides of the nose and mouth are the same width and angle, and the jaw and cheekbone on the left land in the same place as their counterparts on the right.
Measured as a score, it lands lower than people expect. Across 17,032 analyses run between January and August 2026, the average symmetry score was 5.62 out of 10 and the median was 5.5. Only 3.3% of faces score 8 or above and 0.9% reach 9. Nobody scores 10, because nobody is symmetrical.
What a symmetry score of 0 to 10 is actually measuring
On FaceRating.ai the score is an estimate, not a measurement. A vision model looks at the photo and judges how closely paired features match across the midline of the face - the eyes and brows, the two sides of the nose and mouth, the jaw and the cheekbones. No reference points are placed and nothing is measured in pixels or millimetres: the number is the model’s assessment of left-right balance in that one photo.
A 10 would mean the model saw no difference at all between the two halves. That does not occur. A 5.5, the median, means the two halves differ by a normal, unremarkable amount. An 8 means the model saw unusually close balance, which 3.3% of faces get.
Two things follow from that. The first is that a symmetry score describes balance, not attractiveness: it does not say whether the asymmetry it saw is appealing, only that it is there. The second is that it depends heavily on the photograph. A turned head, uneven light or a lopsided expression all change what the model sees, so the same face can score differently in two photos. Our research page explains what a symmetry score means and how to take a fair comparison photo.
Facial symmetry is not the golden ratio
These two ideas get used interchangeably in almost every online discussion of faces, and they are not the same thing at all.
Symmetry is a left-to-right comparison. It asks whether your left side matches your right side, and it says nothing whatsoever about the shape either side has. A face with a very long nose and a very narrow jaw can be perfectly symmetrical, because both halves have the same long nose and the same narrow jaw.
The golden ratio is a top-to-bottom and part-to-part comparison. It asks whether the distances between features fall near 1.618, the classical proportion: the width of the face against its length, the position of the eyes against the height of the head, the thirds of the face against each other. It says nothing about whether the two sides match.
So a face can score high on one and low on the other, and plenty do. They are independent questions asked about the same photograph.
Which of the two actually predicts a rating
This is where the data is unusually decisive, and it is the single most useful thing on this page.
| Measurement | Average across 17,032 faces | Correlation with overall rating |
|---|---|---|
| Symmetry | 5.62 out of 10 | 0.95 |
| Golden-ratio compliance | 77.28% | 0.67 |
Symmetry correlates with the overall face rating at 0.95, the strongest relationship measured anywhere in the dataset. Golden-ratio compliance correlates at 0.67. Both are real relationships, but the difference between them is large: symmetry moves almost in lockstep with the final score, while proportion tracks it loosely.
The group averages make the same point. People scoring 8 or above overall average 8.54 on symmetry, while people scoring under 5 overall average 4.74. That is a gap of 3.81 points on a ten-point scale, opening up purely as a function of how well the two halves match.
Golden-ratio compliance, meanwhile, averages 77.28% with a median of 78% and a maximum of 96%. Nobody in 17,032 faces reaches the ratio. It is a benchmark that everyone falls short of by roughly the same amount, which is part of why it discriminates so poorly. If you want to see your own figure, the golden ratio face test returns it, but it is worth knowing before you run it that it is the weaker of the two signals.
So of the two ideas people repeat constantly about faces, one predicts and the other mostly does not. That is a reasonable thing to update on.
What is the average facial symmetry score?
The direct answer: 5.62, with a median of 5.5. Here is the rest of the distribution.
| Symmetry benchmark | Value across 17,032 analyses |
|---|---|
| Mean score | 5.62 |
| Median score | 5.5 |
| Share scoring 8 or above | 3.3% |
| Share scoring 9 or above | 0.9% |
| Share scoring below 5 | 11.0% |
| Share scoring 7 or above | 11.2% |
| Correlation with overall rating | 0.95 |
Read the bottom half of that table together. 11.0% of people score below 5 and 11.2% score 7 or above, which means the distribution is close to balanced around a middle that almost everyone occupies. Scoring below 5 does not put you at the bottom of anything. It puts you slightly below the median alongside roughly a ninth of everybody. Our piece on how much facial asymmetry is normal covers that distribution and where the bulk of results land.
Why symmetry correlates so strongly, and what that does not mean
A correlation of 0.95 invites a bad interpretation, which is that asymmetry is being punished. That is not what the number says.
Symmetry is best understood as a summary statistic for facial structure in general. The processes that produce a well-proportioned jaw tend to produce it on both sides; the ones that produce an uneven eye line tend to show up elsewhere too. So symmetry moves with everything else the model scores rather than acting on the score independently. A 5.5 on symmetry alongside a 5.5 overall is not two separate findings. It is one finding described twice.
It is also worth being clear about scale. Differences that shift a symmetry score by a few tenths of a point are small enough to come from the photo itself - a slight turn of the head or a change of light - and they are not what other people notice when they look at you. Mirroring one half of a real face onto the other produces something that reads as artificial rather than beautiful. Perfect symmetry is not the target, and it is not what high scorers have.
Getting a reading that is about your face
Symmetry is the score most easily ruined by the camera, because the test assumes you are facing forward and a photograph rarely obliges. Rotate your head five degrees and the far side of your face is genuinely smaller in the image. The model is reading the photo correctly; the conclusion is not. The same applies to head tilt, to a camera held above or below eye level, to an asymmetric smile and to side lighting that leaves one cheek in shadow.
Four fixes handle nearly all of it. Shoot square to the camera, looking at a point directly ahead rather than at your own preview. Keep the lens at eye level. Use a neutral expression, since almost everyone smiles unevenly. And take three photos, because if three readings taken thirty seconds apart come back half a point apart, that spread is telling you how much noise is in the reading rather than how uneven your face is.
The face symmetry test scores each photo independently, so comparing three is straightforward. Then read the result against the table above rather than against 10: 5.5 is the median, 6.5 is above average, 8 is the top 3.3%, and 4.5 is exactly as far below the middle as 6.5 is above it. The full symmetry figures, including the correlation values and the golden-ratio comparison, are in the facial symmetry study.