AI Attractiveness Test
How attractive am I? Upload a photo for a face rating out of 10 and see how many people your age score lower. Free, private, no account.
Example result (AI-generated sample face)
Your score
6.9 / 10
Top 17% of women
One photo is accurate to about ±1.8. A different photo of you can score higher or lower.
Symmetry 59Face contour 60Components
- Overall harmony (Highest)
- 66
- Symmetry
- 59
- Face contour
- 60
- Feature proportions
- 56
- Softness (Lowest)
- 39
Measured separately. Your score is not their sum.
How the attractiveness test works
1. Upload a photo
Pick a photo from your camera roll, or take one with your camera. The attractiveness test analyses it on your device and does not upload it. One photo gives a result in seconds.
2. Check who you are compared with
The attractiveness test compares you with adults of the same gender and adjusts for age. It works both out from the photo, so there is nothing to fill in, and you can change them under your result. Age matters more than most people expect: left uncorrected, a rating model scores people in their fifties about two points lower than people in their twenties.
3. Get your attractiveness rating
You get a face rating from 1 to 10, the percentage of your group that scores lower, and a range showing how far the number could move with a different photo. Below it are component scores for overall harmony, symmetry, face contour, feature proportions and, for women, softness, each with a one-line explanation, plus a share card.
What your attractiveness score means
The scale is centered on your own group. 5.5 means average for your gender and age.
| Score | Higher than | How to read it |
|---|---|---|
| 8.5 | about 98% | Rare. Roughly one person in fifty. |
| 7.0 | about 84% | Clearly above average. |
| 5.5 | 50% | Average for your group. |
| 4.0 | about 16% | Below average in this photo. |
| 2.5 | about 2% | Usually a problem with the photo. Try another. |
Because the scale is spread out this way, it does not hand everyone a 7. Half of all results are below 5.5 by design. If you have used tools where every face seems to score 70 or more out of 100, expect a lower and more honest number here.
How the face rating is calculated
The face rating comes from a small neural network that runs in your browser. We trained it on about 4,600 portraits from a public research dataset (FairFace), each scored for attractiveness from 0 to 100 by a large vision model. We did not take those scores on trust. On 102 studio portraits rated by 2,513 people (the Face Research Lab London Set), the large model's ranking agrees with the human average at 0.8, on a measure where 0 is no relation and 1 is perfect agreement. The small network that learned from it reaches about 0.6 on the same test, which is the level of one typical human rater.
Your raw face rating is then placed inside a reference group of 8,400 adult faces, 600 for each combination of gender and seven ethnic groups, and adjusted for age group. The result is turned into the 1 to 10 scale and a percentile. Nothing in this chain is random. Upload the same file tomorrow and the number will be identical.
Rate my face again: why a second photo scores differently
When we scored several pictures of the same people, one person's results varied by about 0.9 points from photo to photo. That puts the 95% range of a single photo at roughly ±1.8. A 5.5 from one photo could sit anywhere from 3.7 to 7.3, so read a single result as a rough position.
If a result surprises you, have the test rate your face again from a different photo: another day, different light, similar distance. Each photo gets its own result and its own range; the page does not combine them. Photos that are turned more than 20 degrees, too small or show more than one face are refused with a reason, because a forced guess on a bad photo comes out more than 2 points too low.
Which photo to use for the test
The attractiveness test works on photos you already have, and a few choices make the result steadier. Face the camera. Keep the camera at arm's length or further, since a phone held close widens the middle of the face. Your face should be at least 80 pixels wide in the image, which any normal portrait or half-body shot meets. Front light from a window beats overhead light. Glasses are fine. Anything covering the face is not, because hidden points are guessed and nothing warns you. Skip beauty filters; they change what is measured.
What an attractiveness test can't tell you
An attractiveness test rates a photograph of a face. It says nothing about your voice, how you move, how you dress or whether people enjoy your company, and those decide a great deal of real-world attraction.
The model also inherits the taste of what it learned from. It prefers younger faces, which is why we compare within age groups. It gives smiling photos about 0.4 more and underexposed photos up to 0.7 more. You are compared with everyone of your gender and age rather than within one ethnicity, and group averages in the reference data differ by up to 0.3 points for men and about 0.65 for women. Human ratings used for validation came from one set of 102 studio portraits of mostly young adults, so the 0.6 figure is best supported for that kind of picture. For East Asian faces there is no public rating dataset we could license, and we could only confirm that the small network agrees with its teacher.
If a number like this affects how you feel about yourself, treat it as what it is: one stranger's quick opinion of one picture.
Your photo is never uploaded
Face detection, the rating model and the measurement lines all run inside your browser. The photo is read from your device into the page and goes no further; there is no upload, no account and nothing for us to store or delete. You can check this yourself by loading the page, switching your connection off and running the attractiveness test offline. Only the optional paid report sends photos to a server, after you agree to it, and those are deleted after 30 days or whenever you choose.
What research says about facial attractiveness
Decades of studies point to a few measurable cues. Faces close to the population average are rated higher. Symmetry helps a little. Features typical of one's sex play a role, more consistently for women's faces than for men's. A widely cited 2006 review by psychologist Gillian Rhodes summarises these effects and finds all of them modest: none comes close to explaining attraction on its own.
Our own test agrees: a face attractiveness analyzer built only from geometric measurements reached 0.35 against human ratings, well below the image model used here.
Go further with the full report
The free attractiveness test covers the score and the component scores with their measurement lines. The full report adds 46 individual measurements, from brow length to jaw angle, six hairstyle try-ons and a before-and-after image made from your own picture, a written analysis, an action list, and a color direction: how light or deep your coloring is and whether it leans warm or cool. It is a single $9.99 payment with a 7-day refund.




Example from an AI-generated sample face
Attractiveness test FAQ
Is this attractiveness test accurate?
The attractiveness test agrees with the average opinion of a large panel about as well as one randomly chosen person does. That makes it useful as a second opinion and too weak to treat as a verdict.
Am I attractive if I score 5.5?
A 5.5 puts you in the middle of people your age and gender in that photo. Whether a particular person finds you attractive depends on far more than a photo scale can see.
Why do other face rating sites give me a different number?
Every site uses its own method and its own scale, and many are tuned to give flattering numbers. Compare percentiles instead of raw scores, and check whether a site gives the same answer when you upload the same photo twice. This one does.
Does smiling change my score?
Yes, slightly. The same person scores about 0.4 higher when smiling. We do not require a neutral face, because half of ordinary portraits are smiling and the reference group includes both.
Is there an attractiveness test for men and one for women?
It is one attractiveness test with two reference groups. Men are compared with men and women with women, and the report text, hairstyles and styling image are produced separately for each.
Can I upload a photo of someone else?
Please use your own photo. The free attractiveness test runs on your device and we never see the image, but the paid report requires you to confirm the photos are of you and that you are 18 or older.
Is the test really free?
The face rating, percentile, components and share card are free with no limit on attempts and no account. You pay only if you want the full report.