What Makes a Face Look Natural Rather Than Done?

The two oldest received ideas in facial aesthetics are symmetry and the golden ratio. Recent large studies have judged each of them differently. Symmetry first.

There is a study that placed 72 landmarks on 1,550 faces from 10 cultures (Brazil · Cameroon · the Czech Republic · Colombia · India · Namibia · Romania · Turkey · the United Kingdom · Vietnam) and analyzed their relationship to attractiveness.

The finding that “faces closer to the average look more attractive” has itself been replicated for more than 30 years since 1990. When multiple faces are composited into an average face, it was rated higher than most of the individual faces that went into it, and the more faces included in the composite, the more attractive it was rated.

The implication for treatment planning is clear. The goal is not to add features but to reduce the ‘departures’ that catch the eye. On this evidence there is little reason to labor over making the two sides perfectly equal; it fits the evidence better to tidy up the atypicality of one side sagging noticeably more, or one area being noticeably more hollow.

Let us write this part plainly. There is no convincing evidence that the golden ratio (1:1.618) is connected to facial beauty. This is not “the evidence is weak” but closer to a received idea that has been academically refuted.

Here is the conclusion of a study that compared the most widely used ‘golden ratio mask’ with real faces using morphometrics (Procrustes analysis, thin-plate spline).

• The mask is inappropriate for non-European populations, particularly sub-Saharan Africans and East Asians.
• What the mask actually describes is not an ideal face but the proportions of a ‘masculinized Caucasian woman’ seen in fashion models.
• Because the general public strongly prefers femininity in female faces, the mask is inappropriate even for European populations.

A 2024 review is more direct. It concludes that “there is currently no evidence to support the use of the golden ratio in planning orthognathic surgery or facial aesthetic · reconstructive surgery.” The review cites studies of orthodontic · orthognathic patients (2001, 2004) and a study of Afro-Caribbean faces (2018) in which there was no relationship between the golden ratio and appearance ratings.

The same review also points out a methodological problem: studies supporting the golden ratio tend to choose measurement points arbitrarily so that the ratio comes out.

The reason we go out of our way to write this is that applying this yardstick to Korean faces is a matter the literature explicitly identifies as inappropriate. The standard for naturalness is not a formula but the person’s own face, and that agrees with the ‘averageness’ evidence in the section above.

This is the most practical evidence in this article. There is an experiment in which wrinkle and topography cues and evenness of skin color were each removed from 50 photographs of women’s faces (ages 40–74), and 200 observers were asked to estimate age.

There is an even more striking experiment. In one study, facial shape and wrinkles were excluded entirely and only a square patch of cheek skin was shown. Cross-polarized imaging removed even the fine texture, leaving purely the ‘evenness of color’.

• Evenness of skin color correlated with looking younger r=0.493, looking healthier r=0.409, attractiveness r=0.395 (all p<0.001).
• Ages estimated from a patch of cheek skin alone ranged from 17.3 years to 53.1 years. A perceptual span of about 30 years arose with no facial shape information whatsoever.

Put these two studies together and it comes to this. People read age from the ‘surface’ of a face more than from its ‘shape’. So an approach such as skin boosters, which addresses the density and texture of the skin, is not “natural because its effect is subtle” but an approach that acts directly on the axis contributing most to perceived age. This is also why Miso Clinic puts weight on boosters.

When we look at a face and feel “something has been done,” what are we actually doing? It is more likely to be a matter of attention than an aesthetic verdict. There are studies that measured this with eye tracking.

In a study that digitally applied various deformities to children’s faces and tracked the gaze of 46 adult observers, time spent looking at the affected area increased significantly (p<0.01) and time spent looking at the eyes decreased (p<0.05) in faces with a deformity. A study of facial palsy likewise confirmed a measurable degree of distraction in observers.

When people look at a face, they look at the eyes by default. If an unexpected area appears somewhere on the face, the gaze is pulled toward it and the attention that should go to the eyes decreases. This may be the substance of the sensations described as “I can’t quite read the expression” or “my eyes keep drifting there while we talk.”

We will state this honestly. The studies above concern secondary cleft lip deformity and facial palsy, and we could not find a study running the same eye-tracking experiment on faces that had received fillers or lifting. So this explanation is a reasonable inference, not a measured finding. The direction of the inference is nonetheless clear — ‘not catching the eye’ is a safer goal than ‘looking prettier’.

We often hear the worry that “Botox makes your expressions disappear.” This too has been measured.

In one study, 52 female participants were filmed expressing emotions before and one month after 25 units of botulinum toxin to the glabella, and 31 observers rated the recordings.

Another study published in 2026 (118 raters) points the same way. Attributions of anger and surprise fell substantially, while attributions of happiness and neutral expressions were stably maintained.

In other words, what was measured is not “expressions disappear” but “the amplitude of the frowning signal decreases, while happiness and emotional communication itself are preserved.” The ‘frozen face’ people worry about is not what was measured at standard upper-face doses. That said, there is no threshold in the literature for how many units it takes to start looking frozen — which is why we set the dose individually after watching how a person uses their expressions.

A great deal of the outcome described as “looking done” comes from the gap between expectation and reality. So studies measuring how much younger someone actually looks to an observer are useful.

Even with surgery on the whole face, the rejuvenation observers perceive is roughly 4 to 7 years. Please set that beside the finding in the previous section that the experimental effect of improving skin texture was about 10 years. No study has compared the two numbers directly, so we cannot say “A is better than B.” What is clear, though, is that the premise ‘the bigger the change, the younger you look’ does not fit the data well.

One further observation joins this. In a pilot study showing before-and-after photographs to 88 professionals and laypeople, the professionals were more accurate at identifying whether a procedure had been done, but there was no significant difference between the two groups in their ratings of aesthetic improvement. Spotting the ‘tell’ is a trained eye; what the people around you see is generally ‘you look good’.

“Let’s not do a lot at once, let’s split it up” is the way many clinicians work, ourselves included. But we should be careful to distinguish whether this is evidence or convention.

To give the conclusion first, it is closer to convention. We could not find a randomized trial directly comparing split treatment with a single large-volume treatment on the outcome of ‘naturalness’. There was a study that randomized the order of treatments (60 patients), but a natural appearance was achieved in 100% of both arms, so it could not discriminate between the two approaches.

There is, however, evidence on the mechanism side.

• In a study using MRI to examine 33 patients who had received hyaluronic acid in the midface, filler was detected in all 33, with persistence spanning 2 to 15 years. That is a large difference from the 3 to 12 months of duration commonly assumed. — However, this was a selected patient group who had a reason to be imaged, so it must not be read as “this happens in X% of people.” “This happens” is the whole of what this study says.
• Expert literature on the mechanism of midfacial overfilling sees the core problem as ‘layer mismatch’. When a stiff material is placed in a superficial compartment, it cannot accommodate muscle movement during smiling and the distortion becomes visible. — This literature is an expert consensus rather than a randomized trial, and manuscript preparation was supported by a filler manufacturer.

So the accurate way to put it is this. “Splitting it up is our practice, and it is supported by imaging evidence that filler persists for a long time and by the layer-mismatch mechanism. But there is as yet no trial that directly compared and measured whether splitting it up produces a more natural result.”

1. We set the goal as ‘reducing departures’ rather than ‘adding’. What predicted attractiveness was averageness, not symmetry. Rather than mechanically matching the two sides, it fits the evidence better to tidy the one place that catches the eye.
2. We look at the skin itself first. About 80% of perceived age was explained by skin texture. Before changing shape, we first check whether there is anything left to gain from density and texture.
3. We do not apply a formula. Morphometrics has identified the golden ratio as inappropriate for East Asians. The reference is the person’s photographs from when they were younger and their face now.
4. We look at the face in motion. The mechanism of overfilling is ‘layer mismatch’, and that shows up when smiling, not in a still photograph.
5. We go in steps. We say plainly that this is convention, but we choose the side that leaves room to go back over the side that is hard to reverse. The imaging evidence that filler persists longer than expected supports this choice.
6. We tell you the size of the change in advance. It is better to begin knowing that even surgery gives about 4 to 7 years by observers’ reckoning.

• The most solid finding — faces that do not depart from the average read as attractive. In the study of 1,550 faces across 10 cultures and in the independent replication, asymmetry did not predict attractiveness.
• The most practical finding — skin texture explains about 80% of the perception of facial age, and smoothing texture and color together lowered perceived age by 15.85 years. A patch of cheek skin alone produces a perceptual span of about 30 years.
• What needs correcting — the ‘facial golden ratio’ has no evidential basis, and has been explicitly identified as inappropriate for East Asians.
• What you can be reassured about — upper-face botulinum toxin reduces the frowning signal, but the ability to recognize emotion was preserved.
• An inference, not a measurement — explaining ‘you can tell’ in terms of gaze being pulled away. The eye-tracking studies were done only in reconstructive settings.
• Convention, not evidence — “splitting it up is more natural.” There is no direct comparative trial.

Gather up the things that have been measured and the conclusion points in one direction. Naturalness is not the result of restraint, but the result of knowing what people actually look at and putting your effort there. And what people looked at was mostly the skin, and a balance that does not catch the eye.

Written and reviewed by Lee Chi-Hak, MD, medical director of Miso Clinic in Daegu, South Korea. Every study cited above is given together with its design, its size and the limitations the authors themselves recorded, and where we could not find data, we have said that we could not find any.