FaceRating.ai Data Study
How Symmetrical Is the Average Face? Data From 17,032 Scans
Based on 17,032 real analyses · January – August 2026 · Updated August 28, 2026
Data updated . Figures published before that date used the pre-December-2025 model scale and are superseded — the current figures on this page come from a single model on a single 1–10 scale, with no pooling across tools. These are model outputs from submitted photos, not a representative population survey or independent measurements of attractiveness. Associations between scores from the same model do not establish cause and effect.
For help with your own photo, visit our face symmetry test and photo guide. This study describes aggregate model outputs.
Key findings
- • The average facial symmetry score is 5.62/10 (median 5.5).
- • Only 3.3% score 8 or above; near-perfect symmetry (9+) is just 0.9%.
- • Symmetry tracks the overall score almost perfectly (r = 0.95) — far more tightly than golden-ratio compliance (r = 0.67).
- • Faces scoring 8+ overall average 8.54 symmetry; those below 5 average just 4.74.
- • Golden-ratio compliance still climbs cleanly with the score — from 74.98% among 4.0 – 4.9 scorers to 89.03% among 8.0 – 8.9 scorers.
- • No face in the dataset exceeds 96% compliance. Mathematical “perfection” does not occur.
Facial symmetry is one of the most-studied ideas in the science of attractiveness — the theory being that symmetric faces signal good health and are perceived as more attractive. We looked at our AI model’s symmetry estimates across 17,032 real analyses on FaceRating.ai to see how symmetrical the average face scores — and whether it lives up to the theory.
How symmetrical is the average face?
Most people cluster in the 4.5–6.5 range. 41% of all faces sit in the 4.5–5.4 band alone, and genuinely high symmetry is rare — only 3.3% score an 8.0 or above.
| Symmetry score band | Share | Distribution |
|---|---|---|
| Below 4.5 | 4.9% | |
| 4.5 – 5.4 | 41% | |
| 5.5 – 6.4 | 30.6% | |
| 6.5 – 7.4 | 16.8% | |
| 7.5 – 8.4 | 4.4% | |
| 8.5 + | 2.3% |
Share of 17,032 analyses per symmetry band. Bands are rounded to the nearest whole score (e.g. "7.5 – 8.4" ≈ an 8), so they differ slightly from the exact thresholds above: 3.3% score exactly 8.0+, and 0.9% score 9.0+.
Does facial symmetry actually matter?
Here's the most interesting result. When we compare each factor to the overall score, symmetry is almost perfectly aligned with it (r = 0.95), while the famous "golden ratio" compliance is much more loosely linked (r = 0.67). In plain terms: in this data, symmetry is by far the stronger driver of a high rating. Faces that score 8 or higher overall average 8.54/10 symmetry; faces below a 5 average just 4.74.
A note on interpretation: these are the AI model's own scores, so this reflects how strongly the model weighs symmetry when it forms an overall judgment — which happens to line up with decades of attractiveness research pointing to symmetry over any single "magic ratio."
What is a good golden ratio face score?
Golden-ratio compliance averages 77.28% (median 78%), and only 4.4% of faces reach 85% or higher. The ceiling is the striking part: the highest compliance recorded across all 17,032 analyses is 96%. Not one face hits a "perfect" 1.618 across every proportion — which is exactly why chasing a single ideal ratio is the wrong mental model.
That said, the ratio is not noise. Compliance rises steadily with the overall score at every step, from 74.98% among people scoring 4.0 – 4.9 to 89.03% among those scoring 8.0 – 8.9 — a spread of 14.0 points with no reversal along the way. The honest reading is that golden-ratio compliance is a real but weak signal: it moves in the right direction, yet it tracks the overall score far more loosely (r = 0.67) than symmetry does (r = 0.95).
| Overall score | Faces | Avg compliance | Golden-ratio fit |
|---|---|---|---|
| 4.0 – 4.9 | 4,230 | 74.98% | |
| 5.0 – 5.9 | 8,450 | 77% | |
| 6.0 – 6.9 | 2,891 | 78.07% | |
| 7.0 – 7.9 | 824 | 82.28% | |
| 8.0 – 8.9 | 524 | 89.03% |
Bars are scaled from 70% rather than 0% so the differences between bands are visible. Bands with fewer than 50 faces (below 4.0, and 9.0+) are omitted rather than shown as unreliable averages. Nothing in the dataset exceeds 96%.
So both things are true at once: a higher-scoring face does tend to sit closer to the golden ratio, and no face is anywhere near "perfect". If you want to understand your own result, symmetry and overall harmony explain far more of it than phi does.
What does your symmetry score mean?
A symmetry score describes how similar the left and right sides of a face appear in one photo, under one method. It is not a diagnosis, and it is not an attractiveness threshold. A higher value means this model saw more left-right balance in that photo. Different tools use different methods, so their numbers can't be compared with each other.
Is a 7 the same as 70% symmetry?
No. Multiplying a score out of ten by ten changes how it looks, not what it means. Without a defined physical measurement and a validation procedure, a percentage shouldn't be read as the fraction of your face that is symmetrical.
Why can two photos of the same face score differently?
Head rotation changes the visible size and position of features. Expression, shadows and camera distance change what the model sees as well; re-uploading the identical file is designed to return the same score. If you compare photos, keep the conditions the same: front-facing, even light, neutral expression. Don't read a small change as a physical improvement.
What should you do with the result?
Use it to check the photo, or as a starting point for styling, not as a target to perfect. The distribution above shows how this model scores real submitted photos, but that sample doesn't set a "normal" score for everyone. A different hairstyle or pair of glasses can change how a photo reads without changing the face itself. For ideas, see the face shape detector and hairstyle recommendations.
Methodology
This study aggregates 17,032 completed face-rating analyses on FaceRating.ai (January – August 2026). Each analysis includes a symmetry sub-score (0–10) and a golden-ratio compliance estimate (0–100%). Correlations are Pearson's r between each factor and the overall score across all analyses. All figures are aggregate and anonymized — no individual results, images, or personal data are included. Scores reflect the AI model's assessment, not an objective measure of attractiveness. The model was recalibrated in December 2025, so this study uses analyses from January 2026 onward only — mixing the two scales would produce figures describing neither.
How symmetrical is your face?
Explore photo analysis and the access options for a full symmetry report.
Keep reading
Cite this study
These figures are ours, from our own data, and you are welcome to quote them in articles, videos or research with attribution and a link.
FaceRating.ai (2026). How Symmetrical Is the Average Face? Data From 17,032 Scans. Analysis of 17,032 AI facial analyses, January - August 2026. https://facerating.ai/research/facial-symmetry-statisticsHow the data was collected
- 17,032 analyses run on FaceRating.ai during January - August 2026. Aggregate and anonymised; no individual result is published, and no image is included in any figure.
- Scores are produced by a vision model on a fixed 1-10 scale, not by human raters. They describe how this model scores faces, which is what a rating tool is actually asked.
- The sample is self-selected. These are people who chose to have their face rated, so it is not a random sample of any population, and the distribution should not be read as one.
Writing something and want a figure we have not published, or a check on a number before you print it? Ask us.