The genetics of facial features
The human face is one of the most genetically programmed regions of our body. Hundreds of genes contribute to shaping the nose, the distance between the eyes, the curve of the cheekbones or the jaw contour. These genes are inherited equally from both parents — but their expression is far from equal.
Each gene exists in multiple variants called alleles. Some are dominant: they are expressed even if the other inherited copy is different. Others are recessive: they only express when both copies are identical. This allelic lottery explains why a child can have their father's eyes and their mother's nose, or look almost exclusively like one of them.
Dominant and recessive genes: concrete examples
Take eye colour: long considered a simple monogenic trait (brown dominant over blue), modern science has revealed it involves around ten genes, including OCA2 and HERC2 on chromosome 15. Similarly, nose shape — width, projection, nasofrontal angle — is controlled by genes like PAX3, DCHS2 and EDAR, identified in large-scale genomic studies (Adhikari et al., Nature Communications, 2016).
A visually striking dominant trait? The chin cleft, transmitted following a dominant pattern. Or eyebrow shape and upper lip thickness, whose heritability exceeds 60% according to twin studies.
Why do we look more like one parent than the other?
Resemblance is not evenly distributed between the two parents. Several mechanisms explain this:
- Dominant alleles from one parent can systematically "override" the recessive alleles from the other across several traits simultaneously.
- Genomic imprinting: some genes are only expressed based on their parental origin. The IGF2 gene, for example, is only expressed from the paternal copy.
- Genetic drift during meiosis sometimes favours an entire chromosomal block from one parent, resulting in a set of traits characteristic of that lineage.
A study from the University of British Columbia (Bressan & Grassi, 2004) showed that newborns were judged as more resembling their father by neutral observers. This early resemblance bias is thought to be an evolutionary mechanism reinforcing paternal investment.
How does AI analyse facial resemblance?
Modern artificial intelligence facial analysis systems — such as the Azure Face API used by Look Like Me — work in several steps:
- Facial landmark detection: the model locates dozens to hundreds of precise points on the face — eye corners, nostril wings, lip commissures, jaw contour.
- Vectorisation: these coordinates and distances are transformed into a high-dimensional numerical vector representing facial identity.
- Similarity calculation: the vectors of the child and each parent are compared using cosine distance or Euclidean distance. The higher the score, the more geometric features the faces share.
This approach captures aspects of resemblance that genuinely reflect genetic inheritance — facial proportions, feature spacing, overall shape — without constituting a DNA analysis.
The parent-child resemblance test: what is it for?
Beyond playful curiosity, an AI-powered parent-child resemblance test helps you discover which parent you look like most and which facial features were passed down by each lineage. This kind of family resemblance analysis also lets you compare siblings to find out who looks most like mum or dad. It can also spark conversations about heredity, cross-generational resemblances, or simply make for a fun family moment.
What science cannot yet predict
Despite advances in genomics, facial resemblance remains difficult to predict precisely. The environment (diet, sun exposure, ageing), epigenetic modifications and complex interactions between thousands of genes make every face unique. Studies like those from the HUMANFACE consortium continue to explore the genome to identify new markers of facial morphology.
Look Like Me offers you a fun, scientifically inspired window into your family heritage — a starting point for curiosity, not a definitive conclusion.