Between 11 March – 6 June 2026, people ran 1,995 facial analyses on our free tools. We aggregated the results — symmetry, overall harmony, and face shape — into the dataset below. It's anonymous, reported only in aggregate, and it turns up a few things that contradict what the looksmaxxing internet will tell you — and we've checked each finding against the peer-reviewed research.
62%
of faces are oval — the most common shape
63.4/100
average overall harmony score
82
highest score recorded (nobody hit 90)
How to read this
1,995 analyses (1,989 with a full score), 11 March – 6 June 2026. Each result comes from facial landmarks on one uploaded photo, scored by our algorithm — not clinical measurement. Users are self-selected (curious about their own faces, skewing young), so the numbers describe this sample, not the whole population. No personal data is used, and each headline finding is checked against peer-reviewed research (cited at the end).
Oval accounts for nearly two-thirds of all faces; square is the rarest at under 2%. Oval is the "balanced" reference shape, so a classifier leans toward it whenever no other shape strongly dominates — which likely inflates its share.
Against the research
Decades of work find that ‘average’, prototypical faces are rated more attractive than most individuals (Langlois & Roggman, 1990), and that averageness and symmetry are modest-but-real, cross-cultural contributors to attractiveness (Rhodes, 2006). Which of six shape buckets you land in barely registers next to that.
Overall harmony averaged 63.4/100 (median 68). The striking part is the ceiling: the single highest score across all 1,995 faces was just 82, and only 0.4% reached 80+. A "perfect" score effectively doesn't happen — the scale is built to be demanding, so an average result is the high 60s, not 90.
Against the research
There is no scientific ‘perfect face’. Attractiveness is multi-factor — averageness, symmetry and sexual dimorphism each contribute modestly (Rhodes, 2006) — which fits a deliberately demanding scale where nearly everyone clusters in the middle and a 90+ essentially never appears.
The median symmetry score was 75/100, but no metric varied more (standard deviation 27.3). The bottom 10% of faces scored below 14; the top 10% above 91. Some of that range is real, and some is photographic — a slightly turned head or uneven light drops the score — which is exactly why near-perfect symmetry is rare and visible asymmetry is completely normal.
75
median symmetry
14
bottom-10% score
91
top-10% score
98
highest recorded
Against the research
Perfect symmetry barely exists in nature. A statistical-shape study of 321 healthy adults found asymmetry is the norm, with the left side dominant in roughly 81–86% of asymmetric cases (Ercan et al., 2008). Symmetry does track with attractiveness, but the effect is modest (Rhodes, 2006) — so a mid-range score is completely normal.
Diamond and oval faces scored highest for harmony, with square and oblong lowest — but the gap among the top four shapes is only about two points, so shape alone is a weak predictor. (Square and oblong had small samples, so treat the bottom two cautiously.)
| Face shape | Mean harmony | Faces |
|---|---|---|
| Diamond | 64.8 | 213 |
| Oval | 64.0 | 1,242 |
| Heart | 63.0 | 105 |
| Round | 62.9 | 355 |
| Oblong | 55.4 | 47 |
| Square | 51.4 | 31 |
It's the looksmaxxing internet's favourite number — but the evidence is thin. We deliberately left raw golden-ratio scores out of this study because our values were too noisy to report honestly, and the research suggests the "1.618 ideal" is oversold in the first place.
Against the research
The popular ‘golden ratio’ / Marquardt phi-mask standard has been formally critiqued: its goodness-of-fit method is faulty, it best matches masculinised white fashion models, and it doesn’t describe most people’s actual preferences (Holland, 2008). Treat any single ‘ideal ratio’ as a rough guide, not a law — which is also why we left our own noisy golden-ratio numbers out of this study.
An average harmony score is in the high 60s, not the 90s. If a tool hands you a 70, that's genuinely above average — recalibrate your expectations.
Visible asymmetry is the norm, not a flaw. Symmetry is the most variable trait we measured, and much of a low score is just camera angle.
Face shape barely moves overall harmony. Chasing a different 'shape' matters far less than proportion, grooming, and body composition.
Compare yourself to the distribution, not an imaginary 10. The realistic gains are the evidence-based fundamentals, not a perfect score.
Across 1,995 faces analysed on The Face Report, the median facial symmetry score was 75 out of 100, with a mean of about 65. Symmetry was by far the most variable trait measured: the middle of the distribution sat in the high 70s, but the bottom 10% scored below 15 and the top 10% above 91. This fits the research — perfect symmetry essentially doesn't exist. A statistical-shape study of 321 healthy adults found facial asymmetry is the norm, with the left side dominant in roughly 81–86% of asymmetric cases (Ercan et al., 2008). Much of a low score also reflects photo conditions, since a slightly turned head lowers it, so visible, normal asymmetry is expected.
Oval is by far the most common detected face shape: 62.3% of the 1,995 analysed faces were classified as oval. Round was a distant second at 17.8%, followed by diamond (10.7%), heart (5.3%), oblong (2.4%), and square — the rarest — at just 1.6%. Because oval is widely treated as the 'balanced' reference shape, a classifier tends to label faces that don't strongly express another shape as oval, which likely inflates its share. It's worth noting that face-shape categories are a styling convention, not a rigorous scientific construct — research links attractiveness far more to averageness and symmetry than to which shape bucket a face falls in (Langlois & Roggman, 1990; Rhodes, 2006).
The average overall harmony score was 63 out of 100, and the median was 68, so anything in the high 60s is genuinely average. Scores above 75 placed a face in roughly the top 10%. The most striking finding is the ceiling: across nearly 2,000 faces, the highest score recorded was 82, and only 0.4% reached 80 or above — no face scored 90+. That matches what the science would predict: there is no single 'ideal' face, and attractiveness is driven by several modest factors at once rather than one perfectible number (Rhodes, 2006).
Not really. The idea that a 1.618 'golden ratio' or the Marquardt phi-mask defines a beautiful face has been formally criticised in the plastic-surgery literature: the method used to test fit is faulty, the mask best matches masculinised white fashion models, and it doesn't reflect the preferences of most people (Holland, 2008). Symmetry and averageness have better evidence as beauty cues, and even those are modest in size (Rhodes, 2006). We excluded our own golden-ratio figures from this study because the raw values were too noisy to report honestly — but the broader point stands: treat the golden ratio as a rough guideline, not a law of attractiveness.
Diamond (mean 64.8) and oval (64.0) faces scored highest for overall harmony, with heart (63.0) and round (62.9) close behind. Oblong (55.4) and square (51.4) scored lowest — though those samples were small (31 and 47 faces), so treat them cautiously. The practical message is that face shape alone is a weak predictor: the gap among the top four shapes is only about two points, and proportion, symmetry, and averageness matter far more than which of the six categories you fall into (Rhodes, 2006).
The figures are aggregate statistics from 1,995 facial analyses run on The Face Report's free tools between 11 March and 6 June 2026. Each analysis maps facial landmarks from a single uploaded photo and computes algorithmic scores for symmetry, proportion, and overall harmony, plus a face-shape classification. The data is anonymous and reported only in aggregate — no individual results, names, or demographic attributes are included. These are self-selected users curious about their own faces (skewing young and male), and the scores are an AI tool's outputs, not clinical or medical measurements, so the numbers describe this sample. Where possible we check them against peer-reviewed research rather than presenting them as standalone truth.
Run the same free analysis these 1,995 faces did — get your symmetry, harmony, and face shape in seconds and see where you land on the distribution. Private, no signup.
Analyze My Face FreeThe fastest way to use this data is to find where you land on it.
Full Face Analysis
FreeYour complete report — symmetry, harmony, ratios, and face shape — measured the same way as this study.
Symmetry Checker
FreeSee your symmetry score and where it falls against the median of 75.
Face Shape Detector
FreeFind out if you're part of the 62% with an oval face — or one of the rarer shapes.
Looksmaxxing: An Honest Guide
What actually moves these numbers, what's overhyped, and where to start.