The Claim and the Evidence
The claim that the golden ratio (approximately 1.618) governs facial attractiveness has been popularised in cosmetic surgery, beauty guides, and viral videos. The idea is that faces with key proportions approximating this ratio are perceived as most beautiful. Like many appealing simple explanations, the reality is more complicated.
Controlled empirical research on the golden ratio and facial attractiveness is mixed at best. While some studies have found modest correlations between certain facial measurements and the ratio, others find no effect, or find that the relationship is confounded by averageness, faces close to the population mean on many dimensions tend to be both attractive and to approximate many ratio targets simultaneously.
What Actually Predicts Attractiveness
The most robust predictors of facial attractiveness in the research literature are: averageness (faces near the mathematical average of the population are more attractive), symmetry (bilateral symmetry is attractive but has a smaller effect than often claimed), and sexual dimorphism (more pronounced sex-typical features are attractive within groups, but with complex cultural variation).
Notably, AI face recognition does not include attractiveness as a dimension. The embedding encodes identity-relevant geometry, the features that distinguish individuals, not beauty-relevant geometry. High-scoring matches reflect structural similarity to the celebrity's bone configuration, not any assessment of attractiveness.