Face symmetry: what it is and how AI measures it
Facial symmetry describes how closely the left and right halves of a face mirror each other across the vertical midline. Every human face is measurably asymmetric — the question is by how much, and where.
What facial symmetry is
Draw a vertical line through the center of the face — through the middle of the forehead, the nose bridge, and the chin. Facial symmetry is the degree to which features on one side of that line match the position, size, and shape of the same features on the other side: eye corners at the same height, mouth corners equally distant from the midline, jaw angles mirroring each other.
No face is perfectly symmetrical. Studies that mirror one half of a face to build a "perfect" composite consistently find the result looks subtly wrong — the small asymmetries of a real face are part of what makes it look alive.
Why symmetry matters
Symmetry is one of the most studied variables in facial-attractiveness research, alongside proportion and averageness. Across cultures, faces rated as more attractive tend to be more symmetric — but the effect is smaller and more nuanced than social media suggests:
- Symmetry correlates with attractiveness ratings, but so do skin quality, proportion, and expression — no single measurement dominates.
- The correlation is strongest for large asymmetries; differences within the normal range are barely perceptible to observers.
- Perfect symmetry is not the optimum: artificially mirrored faces are rated as less natural and often less attractive than the original.
How Refrakt's AI measures symmetry
Refrakt maps 468 facial landmarks from a single scan, then measures symmetry in three steps:
- Midline estimation. A vertical axis is fitted through midline landmarks — glabella, nose bridge, philtrum, and chin — rather than simply splitting the image in half, so head tilt doesn't distort the result.
- Paired-landmark comparison. Each left-side landmark is compared with its right-side counterpart: distance from the midline, vertical height, and depth. Differences are recorded per region (eyes, brows, nose, mouth, jaw).
- Region scoring. Differences are normalized by face size and summarized per region, so the analysis can say where asymmetry sits (for example, one mouth corner 2 mm higher) instead of producing a single opaque number.
What the research says
Key findings from the scientific literature on facial symmetry:
| Finding | Evidence |
|---|---|
| Symmetry correlates with rated attractiveness | Meta-analyses of perception studies across cultures report a consistent positive correlation of moderate size. |
| Normal faces are asymmetric | Anthropometric surveys find measurable left–right differences in virtually all adults, commonly a few millimeters at the jaw and mouth. |
| Perfect symmetry looks unnatural | Mirrored-composite studies find artificially symmetric faces rated less natural than originals. |
| Self-perception is unreliable | Perspective distortion and mirror habituation make people overestimate their own asymmetry. |
Frequently asked questions
Is a perfectly symmetrical face more attractive?
Not necessarily. Symmetry correlates with attractiveness, but perfectly mirrored faces look uncanny rather than beautiful. Every real face — including famously attractive ones — has measurable asymmetry.
How is facial symmetry measured?
By comparing paired landmarks (eye corners, mouth corners, jaw points) across the facial midline. Smaller positional differences between pairs mean higher symmetry.
Why does my face look asymmetrical in selfies?
Close-range front cameras introduce perspective distortion, and mirrored selfies flip the face you're used to seeing. Both effects make normal asymmetry look larger than it is.
Can facial symmetry be improved?
Minor soft-tissue asymmetries can change with posture, dental work, or targeted procedures; skeletal asymmetry generally requires surgery. Most measured asymmetries are within the normal range and invisible to others.
See your own measurements
Refrakt analyzes your facial symmetry from one scan and shows you exactly where every number comes from — honestly, privately, and without a filter.
Request early access