What Makes a Face Attractive? The Science Behind Facial Attractiveness

Research in facial attractiveness consistently identifies four main structural and perceptual factors: symmetry (bilateral balance between left and right facial halves), averageness (how closely a face's proportions match the population mean), sexual dimorphism (the degree to which facial features express gender-typical characteristics), and expression quality (how warm, genuine, and emotionally open the expression appears).

These four factors have different magnitudes of effect and different degrees of controllability. Symmetry has modest effects that are real but often overstated. Averageness has larger effects that are underappreciated. Sexual dimorphism has complex effects that depend heavily on context. Expression quality has the largest dynamic effects and is the most controllable.

Understanding the relative magnitudes is important because most people trying to improve their attractiveness focus on symmetry and structural features while overlooking the higher-impact expression and contextual factors.

Symmetry preferences are real and cross-cultural. Studies using artificially symmetrised face composites consistently find that more symmetrical versions are rated slightly more attractive. The evolutionary hypothesis is that symmetry signals developmental stability — fewer genetic or environmental disruptions during growth.

However, the effect size is consistently modest in real-world attractiveness ratings — much smaller than popular media coverage suggests. People with highly symmetrical faces are not dramatically more attractive than people with moderate asymmetry. The preference for symmetry is strongest in brief, context-free exposures; in sustained social interaction, expression and warmth signals dominate.

Artificially perfect symmetry can also produce an uncanny effect: faces that are too symmetrical look subtly odd, as though produced by a mirror rather than natural development. Natural faces have small, authentic asymmetries that actually contribute to their appeal — particularly in expressions, where genuine Duchenne smiles are always slightly asymmetric.

Averageness in facial attractiveness refers to how closely a face's proportions approximate the mathematical average of the population. Counter-intuitively, faces closer to the population average are consistently rated as more attractive than faces with highly distinctive features — even when those distinctive features are considered beautiful in isolation.

The mechanism is partly perceptual fluency: average faces are processed more easily by the visual system, and processing ease is experienced as pleasantness. Average faces are also associated with genetic diversity (a large genetic pool tends to produce more average phenotypes) and lower pathogen load — both historically relevant attractiveness signals.

Averageness explains why computer-generated composite faces — created by averaging many individual faces — consistently receive high attractiveness ratings despite being no one's face in particular. It also explains why the most distinctive or unusual features that people think of as 'unique' often do not increase attractiveness ratings in empirical studies.

Expression quality — particularly smile quality — has the largest dynamic effect on perceived attractiveness of any factor measured. A genuine Duchenne smile (with full orbicularis oculi engagement) increases attractiveness ratings significantly more than any static structural feature change. In photo-based studies, the same person smiling genuinely is consistently rated substantially more attractive than the same person with a neutral expression.

The mechanism is multi-factorial: a genuine smile signals positive emotional state (health and wellbeing indicator), warmth and social engagement (approachability signal), and genuine interest in the observer (affiliation signal). All three are attractive signals independently; their combination in a genuine smile produces a powerful aggregate effect.

This is why a mediocre-looking person with a genuinely warm, expressive smile is often experienced as more attractive than a structurally beautiful person with a cold or performed expression. The expression signal overrides the structural assessment in many real-world contexts.

Sexual dimorphism refers to the degree to which facial features express gender-typical characteristics. In males, this means features associated with testosterone development: stronger jaw, prominent brow ridge, wider face. In females, it means features associated with oestrogen development: larger eyes relative to face size, fuller lips, softer jaw contour.

Preferences for dimorphism are context-dependent and not universal. High masculine dimorphism in male faces is associated with dominance and physical health, but also perceived aggression — which reduces attractiveness in contexts prioritising cooperativeness. Studies find that female preferences for male facial masculinity vary with hormonal status, cultural context, and relationship goals.

For practical purposes, sexual dimorphism is a structural factor that is difficult to alter significantly. Its effect on attractiveness is smaller and more context-dependent than the dynamic expression factors above.

AI facial analysis provides objective, quantified data on the expression signals that drive attractiveness ratings. Smile Tracker's blendshape analysis reads the specific muscle activations — eye squint, cheek lift, mouth smile asymmetry — that characterise a genuine versus posed smile. This gives you direct feedback on which expression signals you are producing and how strongly.

Comparing Smile Scores across multiple photos taken under different conditions gives you a precise map of which factors produce the most attractive-reading results for your specific face. Most people find that the score variance across their photos is larger than expected — the same face produces dramatically different readings based on expression quality and lighting.

The blendshape values that most strongly predict high attractiveness scores are the genuine smile indicators: cheekSquintLeft, cheekSquintRight, eyeSquintLeft, eyeSquintRight (orbicularis oculi engagement), and mouthSmileLeft, mouthSmileRight at high balanced values. These are all trainable.

Structural features — symmetry, proportions, dimorphism — are largely fixed. They can be altered surgically, but the effect sizes are modest relative to the effort and risk, and they do not address the expression quality factors that have larger attractiveness impacts.

What you can change, significantly and measurably: expression quality (developing genuine Duchenne smile through daily practice), lighting conditions (facing a window vs. overhead light), posture and framing (chin forward, upright posture), eye engagement (building orbicularis oculi control), and grooming (signalling health and self-care).

Among these, expression development has the largest and most lasting impact — because it improves attractiveness not just in photos but in every social interaction. A genuine, warm, engaged smile is the single most powerful tool for looking more attractive, and it is one of the most trainable skills on this list.