Are face rating apps accurate? What they measure, and what they miss
Face-rating apps are consistent, which is not the same as accurate. What the score is actually computed from, why it is measuring your phone as much as your face, the half of attractiveness no number can hold, and what to ask instead.
Updated 15 September 2026 · 4 min read
What the score is made of
Open a face-rating app and it will find your face, place a few dozen landmarks — the corners of the eyes, the tip and base of the nose, the edges of the jaw, the hairline — and do one of two things with them:
- Compare the ratios between landmarks with a template of proportions the app treats as ideal (the "golden ratio" and its descendants, or averages from a reference set), and turn the distance from that template into a score.
- Run the cropped face through a model trained on a large set of faces that other people rated, and output what that model predicts the rating would be.
Some do both, and most add sub-scores for named features — jawline, cheekbones, eyes, skin — which are the same landmarks grouped by region.
That is the whole mechanism. It is not mysterious, and it is genuinely consistent: give the same app the same photo twice and you get the same number. That consistency is what people experience as accuracy. But consistency only tells you the app is measuring something reliably. The questions are what, and whether it is your face.
Problem one: it is measuring the photo
The app gets one photo, and almost always it is a selfie taken at arm's length. At that distance — roughly 30 cm — the geometry of a face changes. A 2018 paper in JAMA Facial Plastic Surgery modelled it and found the base of the nose appears about 30% wider relative to the face than it would from 1.5 metres, the tip about 7% wider, the face narrower at the sides.
Every one of those changes lands on the landmarks the app measures. Nose width to face width, eye spacing to face width, jaw width — all shift with camera distance. So the score is, in part, a score of how far away you held the phone. Two photos of the same face at 30 cm and 1.5 m will get different numbers from the same app, and nothing about the face changed.
A separate study in 2012 showed the same effect on human raters: faces photographed close were rated less attractive and less trustworthy than the same faces from further away, and the raters could not say why.
Problem two: it is measuring half of the thing
Suppose the photo were perfect. What is the score a score of?
The research on how people judge faces gives a clear and slightly awkward answer. There is a shared component — a large meta-analysis in 2000 found raters agree above chance, within and across cultures, on which faces are more attractive, with averageness and symmetry among the cues. And there is a personal component. A 2006 study that separated the two found that roughly half of the variation in attractiveness judgements is private taste: things one rater responds to and another does not.
A single number, by construction, can only describe the shared half. It cannot say anything about the half where the person looking at you has their own taste — which, for any specific person's reaction, is the half that decides.
Problem three: it misses what actually moves
Sit across from someone for the first time and what registers, in order, is roughly: the outline (hair, clothes, posture), the face as lit, the expression, and only then the proportions. Almost all of that is changeable and none of it is in the score:
- Hair — the cut changes the outline of the head, which changes how the face reads. Not measured.
- Clothes and fit — the second-biggest part of the outline. Not measured.
- Light — moves the apparent shape of the eye sockets, jaw and cheeks. Actively confuses the landmarks.
- Expression — a real smile changes how the same face is judged more than most bone measurements. Some apps penalise it because it moves the landmarks.
- Grooming — brows, beard line. Partly measured, in the wrong direction (a beard that improves the jaw line can lower a jaw score).
- Colour near the face — not measured.
So the number leaves out the levers and reports on the fixed part, from a distorted photo, describing half the phenomenon. That is a lot of caveats for a score people take personally.
What the apps get right
They are honest about being a number, and the number is repeatable. For someone who wants a scoreboard — check, change something, check again — that is a real, if narrow, use. Some of the AI-trained ones do pick up on things like skin evenness and eye openness that reflect real perception. And they are cheap to try, even if the subscriptions add up: several popular ones charge around $3.99 a week to see the result after a free scan.
The problem is not that the number exists. It is that it gets mistaken for a fact about you, and then for a plan.
The better question
"What is my score?" has no actionable answer. "What should I change first?" does, and it is a question a number cannot answer but a person looking at your photos can.
That is the whole premise of Auraspire's face scan: one photo, and you get your face shape, your face type, how a stranger is likely to read you in the first second, and the changes that would shift it. No score, no percentile, nothing about weight, height or anything medical. The first impression is an estimate, and it says so, because certainty is the thing this page has just spent a thousand words explaining the limits of.
Whatever you use, take the photo from two metres away first. The selfie guide explains why that alone changes the answer.
Questions people ask
- Do different apps give different scores for the same face?
- Usually, yes — often by a full point or more on a ten-point scale — because they use different landmark sets, templates and training data. Each app is consistent with itself, which feels like accuracy, but the disagreement between apps is the tell that no one of them is measuring a fact.
- Are the apps that use AI trained on real ratings better?
- Better than a fixed geometric template, in that they capture what a group of raters tended to like. But they inherit that group's tastes and biases, they are still fed a close-up selfie, and they still output one number for something that is roughly half personal. They also cannot see your hair, clothes or the light — or if they can, they do not know what to do with them.
- Can a score motivate me to improve?
- For some people it does, briefly. The risk is that a number sticks in the head as a fact about you rather than a fact about a photo run through a template. If you use one, use it once, from a photo taken two metres away, and then put it down.
- Is there such a thing as an objectively attractive face?
- There is a shared component — raters across cultures agree more than chance on averageness, symmetry and a few other cues — and a large personal component. Both are real. Apps measure only the shared part, and only the part of it that survives a distorted photo.
Sources
- Ward B, Ward M, Fried O, Paskhover B. Nasal Distortion in Short-Distance Photographs: The Selfie Effect. JAMA Facial Plastic Surgery, 2018. — nasal base ~30% wider at 30 cm than at 1.5 m, from a model built on average facial measurements
- Bryan R, Perona P, Adolphs R. Perspective Distortion from Interpersonal Distance Is an Implicit Visual Cue for Social Judgments of Faces. PLoS ONE, 2012. — the same face photographed close was rated less attractive and less trustworthy than from further away
- Hönekopp J. Once more: is beauty in the eye of the beholder? Relative contributions of private and shared taste to judgments of facial attractiveness. Journal of Experimental Psychology: Human Perception and Performance, 2006;32(2):199–209. — roughly half of the variance in attractiveness judgements is private taste
- Langlois JH, Kalakanis L, Rubenstein AJ, et al. Maxims or myths of beauty? A meta-analytic and theoretical review. Psychological Bulletin, 2000;126(3):390–423. — raters agree above chance within and across cultures — the shared component is real