An AI attractiveness test scores a face the way people rate faces. This one runs a model trained on 224,114 ratings of 3,897 faces on your device and returns a score out of 100. It then swaps each part of your face for a typical one to show what lifts the number and what pulls it down.
How attractive am I on a scale of 1 to 10?
Divide your score by 10: a 74 is a 7.4 out of 10 and lands in the Above average band. The bands come from the same model, so a 7.4 means the same thing on every photo.
| Score | Out of 10 | Band | Share of 827 test faces |
|---|---|---|---|
| 90-100 | 9 to 10 | Top range | 2% |
| 75-89 | 7.5 to 8.9 | High | 15% |
| 60-74 | 6 to 7.4 | Above average | 34% |
| 40-59 | 4 to 5.9 | Average | 40% |
| 0-39 | 0 to 3.9 | Below average | 9% |
What does the face rating AI measure?
The face rating AI reads your face the way people do, as a whole, then breaks the score into five parts: eyes, nose, jawline, skin and hair.
478 points
478It maps 478 points on your face to find each part and line the face up straight. This runs in your browser, so the photo stays on your device.
One score from 224,114 ratings
0.92A model trained on 224,114 ratings of 3,897 faces reads the whole face at once. On 2,200 faces it had never seen, its scores matched the average human rating at 0.92.
Eyes
+2.9°Canthal tilt, spacing and the brow line. The test swaps your eyes for typical ones and scores again: the gap is what your eyes are worth.
Nose
1.00×Width and length against the rest of your face, tested the same way.
Jawline
130.7°Jaw angle and chin, the part looksmaxxing talks about most. The test shows whether yours adds points or costs them.
Skin
+19Texture and evenness. Shadows and soft focus read as texture, so light matters most here.
Hair
+2Hair and how it frames your face. The one part you can change by next week.
Whether you call it a face rater, a rate my face tool, a beauty test or a pretty scale test, it works the same way here: one photo in, one score out, and the parts behind it.
Does symmetry or the golden ratio decide your score?
No. We tested a geometry formula built on symmetry, facial thirds and fifths, length to width, eye tilt and jaw angle against 2,200 faces rated by people, and its scores ran slightly against their ratings, a correlation of −0.20. The model on this page, trained on the ratings themselves, matched the same people at 0.92. One person matches the rest of the crowd at 0.78.
Golden ratio calculators score a face against fixed proportions in the same way. Proportions still show up in your result: canthal tilt, jaw angle and nose width come from the same 478 points, and the model decides how much each one matters on your face.
- Geometry formula−0.20
- One person vs the crowd0.78
- This model0.92
How do I raise my attractiveness score?
Start with the photo: face the camera, hold it at eye level and arm's length, and use even light from the front.
A turned head or a lens too close changes how the face reads. Straight on, eye level, arm's length.
Even light from the front. Side light and shadows read as texture on the skin.
In our test on 153 faces, a closed-mouth smile moved the score by 3.5 points on average. Try both and compare.
The parts you can actually change. Your full read picks the haircut for your face shape, the beard line and the first skin fix.
Can ChatGPT rate my face?
It can describe your face, but its rating moves with the wording: people report the same photo landing at 6, then 7.5, then 8, then 5. This test is consistent. The same fixed model scores every photo, so the same photo gets the same number every time.
Run the scanAm I ugly if my score is low?
A low score means this photo rates below most of the 3,897 faces the model learned from. Your result also shows which part costs you the most, and that is where to start.
Skin, hair and beard are the parts you can change, and a straight-on photo in even light shows your face at its best. Then scan a second photo and compare the two.
Frequently asked questions
Yes. Drop a photo and you get your score, the part costing you the most and the score for each part, with no account and no quiz. The scan runs on your device. RIZZER Pro adds the full read: what reads in the first three seconds and what to change first.
The free scan never uploads it: the model runs in your browser, and RIZZER receives only the numbers, your score and the part scores, for the comparison base. The photo goes to our server only if you order the full read in RIZZER Pro.
It finds 478 points on your face and lines the face up, then a model trained on 224,114 ratings of 3,897 faces gives it a score out of 100. After that it swaps each part for a typical one and scores again to show what that part adds or costs.
On 2,200 faces it had never seen, its scores matched the average human rating at a 0.92 correlation, while a typical single rater matches the crowd at 0.78. The photo matters too: a straight-on, evenly lit frame gives the most comparable score.
Yes. An 8 out of 10 is an 80 here and sits in the High band, which 17% of test faces reach. The High band covers every score from 75 to 89. Your result shows the parts that put you there and the one costing you the most.
Anything from 60 up sits above the middle of the scale: 60 to 74 is Above average, 75 to 89 High, 90 and up Top range. Across 827 test faces, 17% scored 75 or more and 2% reached 90. The median face lands near 60.
It can tell how people are likely to rate a photo of you. This model learned from 224,114 ratings and matches the average rating at 0.92 on faces it never saw, closer than one person agrees with the crowd (0.78). It reads the photo, so light, angle and expression move the number too.
Scan a straight-on photo and read two things: the score, where 60 and up sits above the middle of the scale, and the part scores, which show what lifts you and what pulls you down. Scan a second photo to see how much the photo itself moves it.
Because it writes a new answer each time, and the wording of your prompt steers it: people see the same photo rated 6, then 7.5, then 8. This test uses one fixed model, so the same photo gets the same score every time you run it.
Check two things under your score. First, the part costing you the most and by how many points. Second, the frame warning: a turned head, a tilted chin, a big smile, low light or a soft photo each change how the face reads. Retake straight on and compare the two scans.
Yes. The model learned from 1,917 men's faces and 1,980 women's faces and scores every face on the same scale from 0 to 100. Men and women get the same parts in the result: eyes, nose, jawline, skin and hair.
Yes, as many as you like. Compare with another photo puts two scans side by side with both scores and the difference for each part, so you can pick the photo that scores higher. Everything runs on your device, and the photos stay there.
People respond to the whole face in the ratings this model learned from: clear skin, defined eyes and brows, a balanced jawline and hair that frames the face. The part scores in your result show which of these works for you and which costs you points.
RIZZER Pro reads your photo the way a match does: what reads in the first three seconds and what to change first, from the haircut for your face shape to the beard line, skin and your best angle. It opens right after your free score.



