Ask ten people what makes a face attractive and you'll get ten answers: good bone structure, nice eyes, a strong jawline, "something about them." Ask the research literature the same question and something surprising happens — the answers converge.
Since the late 1980s, psychologists have run hundreds of controlled studies on facial attractiveness, using computer-morphed faces, cross-cultural rating panels, infant looking-time experiments, and twin studies. The picture that emerges is not a list of features. It's a small set of structural properties that predict attractiveness ratings with remarkable consistency — plus one finding that complicates the whole story in an interesting way.
Here's what the evidence actually shows.
1. Averageness: the most counterintuitive finding in the field
In 1990, Judith Langlois and Lori Roggman published a study in Psychological Science with a title that sounds like an insult: "Attractive Faces Are Only Average."
Their method was elegant. They digitized samples of male and female faces, mathematically averaged them into composite images, and asked adults to rate the attractiveness of both the individual faces and the computer-generated composites. Both the male and female composite faces were judged more attractive than almost all of the individual faces that went into them — and a strong linear trend showed the composites became more attractive as more faces were averaged together.
This is the "averageness effect," and it's one of the most replicated results in the field. It's also widely misunderstood. "Average" here doesn't mean unremarkable — it means mathematically central. A face close to the population mean has no unusual proportions: nothing sits too far apart, too close together, too large, or too small relative to everything else. Faces like that are rare. Almost nobody is actually average.
Why would the brain reward this? Two explanations compete, and both may be right. The evolutionary account says proximity to the population mean signals genetic health and developmental stability. The cognitive account says our brains build prototypes from everything we see and simply process prototype-like stimuli more fluently — and mistake that fluency for beauty. Notably, the averageness effect isn't limited to faces; a strong relationship between averageness and attractiveness has been found for dogs, wristwatches, and birds too, which is hard to explain with mate choice alone.
Curious where your own proportions sit relative to the population mean? That's precisely what our attractiveness test measures.
2. Symmetry: real, replicated, but smaller than you think
Facial symmetry is the trait most people have heard of, and the evidence supports it — with caveats that rarely make it into popular articles.
In her 2006 review in the Annual Review of Psychology, Gillian Rhodes examined the full body of evidence on biologically-based beauty standards. Her critical review and meta-analyses found that averageness, symmetry, and sexual dimorphism are all attractive in both male and female faces, and across cultures. That cross-cultural consistency matters: the early emergence of these preferences in development, combined with cross-cultural agreement, challenges the long-held view that attractiveness standards are arbitrary cultural conventions.
So symmetry is real. But here's the caveat: the asymmetries that drive rating differences in experiments are often larger than the ones present in ordinary faces. Some research suggests that the low degree of facial asymmetry found in normal people may not meaningfully affect attractiveness ratings, likely because observers simply aren't tuned to perceive asymmetries that subtle.
The practical translation: pronounced asymmetry is noticeable and does affect perception. The millimetre-level asymmetry that nearly every human face has — including every face you consider beautiful — is largely invisible to observers. If you've ever flipped a selfie horizontally and felt something was "off," you were detecting a change from a familiar image, not discovering a flaw.
You can measure your own symmetry precisely with our face symmetry test.
3. Sexual dimorphism: the trait with the most variation
Sexual dimorphism refers to how strongly a face expresses typically masculine or typically feminine structural cues — brow ridge prominence, jaw width, cheekbone position, lip fullness, facial adiposity.
Rhodes's meta-analytic review found sexual dimorphism attractive across cultures alongside averageness and symmetry. But of the three, this is where preferences vary most — between individuals, between cultures, and even within the same person over time. Preference for masculinity in male faces is notably unstable across studies, shifting with the rater's own characteristics, relationship context, and what's being judged (short-term attraction versus long-term partnership).
This is where a lot of online "looksmaxxing" content overreaches, treating a single masculinity or femininity ideal as an objective target. The literature doesn't support that. It supports dimorphism mattering — with the direction and magnitude of the effect varying considerably.
4. Skin quality: underrated and unusually actionable
Beyond structure, skin texture and colouration independently influence attractiveness judgments. In their 2011 review in Philosophical Transactions of the Royal Society B, Little, Jones and DeBruine catalogued the characteristics that shape face preferences: symmetry, sexually dimorphic shape cues, averageness, skin colour and texture, and cues to personality.
Skin deserves more attention than it gets in these discussions for a simple reason: unlike bone structure, it's substantially modifiable. Sleep, hydration, sun protection, and dermatological care all measurably change skin appearance. If you're looking for the highest-leverage change available to most people, it isn't in the skeleton.
5. The caveat that changes everything: individual preference is mostly environmental
Here's the finding that most articles on this topic omit — and it may be the single most important one.
In 2015, Laura Germine and colleagues published a twin study in Current Biology investigating why people disagree about attractiveness. Their starting point was that although certain characteristics are broadly considered more attractive — symmetry, averageness — people routinely disagree with each other about the relative attractiveness of specific faces. Using a twin design, they separated genetic from environmental contributions to individual preference.
The result was striking. Almost all reliable variation in face recognition ability was attributable to genes — but most reliable variation in individual face preference was attributable to environment. And not shared environment, either. It's not the school you attended or your parents' income — it's the experiences unique to you: the faces you've encountered in media, your daily social interactions, perhaps even the face of a first partner.
The researchers ruled out measurement error as an explanation, concluding that individual aesthetic face preferences really are shaped primarily by individual life experiences.
This produces the honest, two-part answer to "what makes a face attractive":
There is a shared layer. Averageness, symmetry, sexual dimorphism, and skin quality shift ratings consistently, across cultures, in a direction most raters agree on. This layer is measurable — it's what any attractiveness scoring system is actually quantifying.
There is a personal layer. On top of the shared layer sits substantial individual variation, driven by each rater's unique history. This layer is not measurable by any algorithm, and it's the reason a person who leaves one room unnoticed can walk into another and stop conversation.
What this means for your score
If you take an attractiveness test — ours or anyone's — what you're getting is an estimate of the shared layer. A well-built system measures proportional relationships, symmetry, and dimorphic cues against population norms, and it can do that consistently.
What it cannot do is predict how any specific person will perceive you. That depends on their unshared environment, which no model has access to.
That's not a limitation worth apologizing for — it's the accurate framing. A score tells you something real and narrow. Treat it as one data point about structural convention, not a verdict about your worth or your romantic prospects. The research is quite clear that the second thing isn't measurable.
Frequently Asked Questions
Is facial attractiveness objective or subjective?
Both, in measurable proportions. Cross-cultural meta-analyses show broad agreement on which faces rate highly, indicating a genuinely shared component. But twin research shows individual preferences are driven mostly by each person's unique environment. Attractiveness has an objective core surrounded by substantial personal variation.
Does symmetry really matter?
Yes, but less dramatically than commonly claimed. Symmetry is a replicated predictor of attractiveness ratings across cultures. However, the subtle asymmetries present in ordinary faces are often below the threshold observers actually notice.
Why are average faces attractive?
Because "average" means proportionally central, not plain. Composite faces built by mathematically averaging many faces reliably outrank the individual faces they're made from — and become more attractive as more faces are added. Explanations include signalling of developmental stability and easier cognitive processing of prototypical stimuli.
Can you change how attractive your face is?
Structural traits like bone proportion are largely fixed, but skin texture and colouration measurably affect attractiveness judgments and respond to sleep, sun protection, hydration, and dermatological care. Photographic factors — lighting, angle, expression — also substantially change how a face is perceived.
How accurate are AI attractiveness tests?
They can measure the shared, structural layer reasonably consistently. They cannot predict individual attraction, because that depends on the viewer's personal history. Read our full breakdown in [How Accurate Are AI Attractiveness Tests?](/blog/how-accurate-are-ai-attractiveness-tests)
References
- Langlois, J. H., & Roggman, L. A. (1990). Attractive Faces Are Only Average. Psychological Science, 1(2), 115–121. https://doi.org/10.1111/j.1467-9280.1990.tb00079.x
- Rhodes, G. (2006). The Evolutionary Psychology of Facial Beauty. Annual Review of Psychology, 57, 199–226. https://doi.org/10.1146/annurev.psych.57.102904.190208
- Little, A. C., Jones, B. C., & DeBruine, L. M. (2011). Facial attractiveness: evolutionary based research. Philosophical Transactions of the Royal Society B, 366(1571), 1638–1659. https://doi.org/10.1098/rstb.2010.0404
- Germine, L., Russell, R., Bronstad, P. M., et al. (2015). Individual Aesthetic Preferences for Faces Are Shaped Mostly by Environments, Not Genes. Current Biology, 25(20), 2684–2689. https://doi.org/10.1016/j.cub.2015.08.048