Golden ratio face: what the evidence actually says

The golden ratio, about 1.618, is often presented as the secret of an attractive face: the closer your proportions come to it, the better you look. The evidence does not support that. Studies that measured real faces against golden-ratio targets found weak or no links with how attractive people rated them, and a 2024 review concluded there is no convincing evidence connecting the golden ratio to ideal human proportions or facial beauty. What research does link to attractiveness is more ordinary: faces close to the average shape, reasonably symmetric, with features typical of their sex. This guide explains where the claim comes from, what the studies found, why 1.618 can be found in any face, and what to make of a golden ratio score.

What the golden ratio is

Split a line into two parts so that the whole line relates to the longer part exactly as the longer part relates to the shorter one. That ratio is the golden ratio, written φ (phi):

It has real mathematical properties. Euclid described it around 300 BC as dividing a line in “extreme and mean ratio”, and the ratios of consecutive Fibonacci numbers (8/5, 13/8, 21/13…) close in on it. It also appears in some patterns of plant growth, such as the spirals of seeds in a sunflower head.

The aesthetic claims came much later. Many popular stories, that the Parthenon, the pyramids or the proportions of the human body are built on φ, do not survive measurement. A well-known paper in a mathematics journal went through these claims one by one and found that most rely on choosing whichever measurements happen to fit, or on rounding generously. The famous 1509 book on the “divine proportion” by Luca Pacioli, illustrated by Leonardo da Vinci, was about geometry; the idea that φ is the key to beauty was mostly added by later writers.

Where the golden ratio face comes from

In facial aesthetics, the idea took hold in the twentieth century. The orthodontist Robert Ricketts proposed in 1982 that attractive faces contain many “divine proportions”. In the 1990s and 2000s the surgeon Stephen Marquardt built a mask out of golden-ratio shapes and promoted it as the archetype of the attractive face. It spread widely, first among cosmetic surgeons and then online, where golden ratio face calculators and celebrity rankings now draw large audiences.

What the evidence says

Few “ideal” ratios predict attractiveness

A Dutch study asked 76 people to rate photos of 64 adolescents, then measured 27 frontal ratios and 26 profile angles that the literature had proposed as ideal, including golden-ratio ones. Only two ratios and three angles related to the ratings at all. Together they explained 28.7% of the variation in ratings, leaving most of it to everything else about the faces.

A study in Hong Kong showed 3D photos of 30 people to 161 raters and tested golden-ratio targets, neoclassical canons and other “ideal” values. More than half of the measurements in the female faces differed significantly from their supposed ideal, and only a few of the ideal parameters were related to how attractive the faces were rated.

The Phi mask fits fashion models, not most preferences

An analysis of Marquardt’s mask using shape statistics found that its proportions best match masculinised white women, the look of many fashion models. It fits non-European faces poorly, especially East Asian and sub-Saharan African faces. And because most people prefer more feminine proportions in female faces, aiming toward the mask would move a face away from what most raters prefer.

A counterpoint, and what it really shows

Not every study is negative. A German study compared magazine cover models and former orthodontic patients rated attractive or not, against twelve golden-ratio distances. All groups deviated considerably from the golden values, but the faces rated less attractive deviated more. The same result is expected if attractive faces are simply closer to average, since a ratio target near the population average will sit closer to average faces too. It does not show that 1.618 is the reason.

The verdict of a recent review

A 2024 review in a maxillofacial surgery journal traced golden-ratio claims from the nineteenth century to the current clinical literature. Its conclusion: there is no convincing evidence that the golden ratio is linked to idealised human proportions or facial beauty, and no evidence to support its use in planning jaw or facial surgery or judging its results.

Why 1.618 shows up in every face

There is a simple reason golden-ratio analyses always find something: a face offers a huge number of ratios to choose from. We tested this on the average face our scan is built on.

Pick a few of those 412 and you can draw a convincing golden-ratio overlay on almost anyone. Pick different ones and you can make the same face look like a poor match. That is why golden-ratio scores from different apps often disagree on the same photo, and why the method needs to be fixed in advance before its results mean anything.

How the average face compares with 1.618

If the golden ratio described faces, you would expect the average face to land on it for the ratios that golden-ratio methods use most. It does not. On MediaPipe’s canonical face, the average face our scan is built on:

• Mouth width to nose width is about 1.38, not 1.618. The nose is about 73% as wide as the mouth.
• The gap between the eyes is about 1.43 times the width of one eye, not 1.618 and not the 1.0 of the classical canon.
• The vertical thirds come out at 32% / 33% / 35%: close to equal, which is the neoclassical canon, not a golden-ratio division.

The “new golden ratios” from the attractiveness research do not involve φ either: 46% and 36% are simply where the average face puts the eyes and mouth. When the target that predicts ratings is the average, a fixed constant like 1.618 can only ever be right by coincidence.

What a fair test would look like

The golden-ratio idea could be tested properly, and the better studies above come close. A fair test would fix the ratios in advance, before seeing any ratings, so they cannot be picked to fit. It would use ordinary faces from the population the claim is about, not fashion models. It would have the faces rated by people who do not know the hypothesis. And it would check whether 1.618 predicts ratings better than the simplest alternative: closeness to the average face. So far, where studies have done some of this, the golden ratio has not beaten averageness, and most “ideal” ratios have not predicted ratings at all.

Why the idea is so persistent

If the evidence is this thin, why is the golden ratio face everywhere? A few reasons. A single number feels scientific and gives a simple answer to a hard question. Overlays of spirals and grids on a famous face look convincing, because, as the chart above shows, some lines will always fit. The claim has been repeated in clinical papers, textbooks and surgery marketing for decades, which makes it look settled. And it is easy to turn into an app. None of that adds evidence; it only explains the reach.

What proportions do matter

The research on facial attractiveness points to three geometric features that hold up across studies and cultures, which a review with meta-analyses by Gillian Rhodes summarised in 2006:

• Averageness. Faces close to the population’s average shape. In 1990 Langlois and Roggman showed that composites made by averaging many faces are rated more attractive than most of the faces that went into them. When Pallett and colleagues searched for the most attractive spacing of the eyes and mouth, they found it at about 46% of face width between the eyes and 36% of face length from eyes to mouth. They called these the new golden ratios, and noted that they match the average face.
• Symmetry. Left and right sides that match closely, though not perfectly. The facial symmetry guide covers how it is measured and how much it matters.
• Sex-typical features. In female faces, more feminine proportions are consistently preferred. In male faces the preference for masculine features is weaker and varies.

Beyond geometry, skin, expression, grooming and body fat all move ratings, often more than any ratio. A face reads as a whole.

How to read a golden ratio score

The golden ratio face test measures your face length to width, eye spacing, nose to mouth width and vertical thirds, and reports how close they come to 1.618 and to the neoclassical canons. It is there because people want to know, and it is an accurate measurement of those ratios on your face. Read it for what it is:

• A description, not a rating. A lower match means some of your ratios sit further from 1.618, not that your face is less attractive.
• Not part of our attractiveness score. The attractiveness test uses symmetry, averageness, proportions, eyes, jaw and profile, as the methodology page documents. The golden ratio is not an input.
• Sensitive to the photo. Camera distance and head tilt change face length and the thirds, so measure at eye level from at least an arm’s length.

If you want the proportions that research does link to how faces are seen, look at your thirds, fifths and eye spacing and your symmetry. The facial thirds and fifths guide explains what real faces measure compared with the classical canons.

One more thing we don’t do: rank celebrities or anyone else by golden-ratio score. Those lists use the same flexible methods described above, on photos taken with different lenses and angles. The free result is your own golden-ratio match; the full report ($14.99) adds your symmetry by region, your averageness and your proportions, drawn on your own scan.