FaceRating.ai Research

    Data studies from 17,032+ real face analyses

    Most of what's written about facial attractiveness online is opinion. We publish the numbers instead: original, aggregate studies built from real, anonymized analyses performed on FaceRating.ai. No individual data, just what the patterns actually show.

    Data study · n = 17,032

    What's the Average Face Rating Score? We Analyzed 17,032 Faces

    The average AI face rating is 5.59 out of 10 — and only about 1 in 12 people score a 7 or above. Here is the full distribution from 17,032 analyses, and what your score actually means in percentile terms.

    Read the study →
    Data study · n = 23,709

    The Most Common Celebrity Lookalikes: 71,000 AI Matches Analyzed

    Across 71,031 AI lookalike matches, two faces come up more than any other: Dev Patel and Timothée Chalamet. Here is the full ranking of the most-matched celebrities — and who scores the strongest resemblance.

    Read the study →
    Data study · n = 17,032

    Potential vs Actual: How Much Can a Face Rating Really Improve?

    Every analysis estimates a potential score alongside the current one. Across 17,032 scans the gap averages just 1.31 points, most people sit between 1 and 2, and not one person under a 5 has a potential of 8 or higher.

    Read the study →
    Data study · n = 17,032

    Which Facial Feature Matters Most? 17,032 Faces Analyzed

    Every analysis scores twelve facial features. The one people most often score worst on is not the nose or the jawline - it is skin, for 38% of faces, and it is the only feature near the top of that list that responds to anything you can do.

    Read the study →
    Data study · n = 17,032

    What Is the Most Common Face Shape? 17,032 Scans Analyzed

    Two shapes cover almost 87% of all faces, and they split sharply by gender: oval is 71% of women, oblong 55% of men. Round scores lowest for everyone. Square scores highest overall - but that turns out to be true only for men.

    Read the study →
    Data study · n = 17,032

    How Symmetrical Is the Average Face? Data From 17,032 Scans

    The average facial symmetry score is 5.6 out of 10, and near-perfect symmetry is rare. But the striking finding is how tightly symmetry tracks the overall score — far more than the "golden ratio" does.

    Read the study →

    Headline findings

    The five numbers most worth knowing, each drawn from the study beside it.

    0.95 vs 0.67
    Facial symmetry correlates with the overall rating at 0.95. Golden-ratio compliance, the far more famous idea, manages only 0.67 on the same 17,032 faces. One of the two predicts how a face is rated. The other mostly does not. Symmetry study
    5.59 / 10
    The average score across 17,032 analyses, with a median of 5.3. Almost half of everyone lands between 5.0 and 5.9, and only 3.3% reach 8 or above. Score distribution study
    38%
    Skin is the weakest feature for 38% of people, more than twice any other. It is also one of only two features that genuinely respond to effort, which makes it the rare case where the most common problem is also the most fixable. Feature study
    2.6%
    Square is the rarest face shape and the highest scoring overall, but that holds for men only. Male square faces average 6.97; female square faces average 5.14, below the female average of 5.71. The female sample is small (n = 53), so treat that half as indicative. Face shape study
    +1.31
    The average gap between a face as scored and its estimated potential is 1.31 points, and it is remarkably consistent: the middle half of people sit between 1 and 1.7. Nobody is one change away from a different face, and almost nobody is stuck either. Potential study

    For journalists and researchers

    Every figure on these pages comes from our own analyses, and you are welcome to quote any of them with attribution and a link. Each study carries its own citation line, its sample size and the window the data covers.

    Two things worth stating plainly, because they change how the numbers should be read. Scores come from a vision model on a fixed 1-10 scale rather than from human raters, so they describe how this model scores faces. And the sample is self-selected: these are people who chose to have their face rated, not a random sample of any population.

    Need a cut we have not published, or a check on a figure before you print it? Get in touch.

    Get your own analysis

    Free AI face rating with a full feature breakdown and celebrity lookalike match.

    Rate my face, FREE