Start here
Why your rating might look wrong (and how to fix it)
3 min read · Updated October 4, 2026
If a score feels way off, the cause is almost always the photo, not your face. Here's how to tell, and how to fix it.
The photo is doing it, not your face
By far the most common reason for a strange result is the photo. A selfie taken close-up on a wide-angle phone lens distorts the centre of your face and throws off nearly every ratio. A tilted head moves the thirds and the canthal tilt. Harsh light fakes or hides bone definition.
Before you believe a bad number, retake the photo (the full checklist is in Taking the right photo): rear camera, about 1.5 metres away, camera at eye level, head level and straight on, soft even light, neutral closed-mouth expression, hair off the forehead.
Landmark detection can slip
The measurements are only as good as the detected landmarks. On a low-resolution, dark, blurry or sharply-angled photo, the model can place points slightly wrong — and a few pixels of error on a key point shifts the ratio it feeds. A clean, well-lit, front-on photo gives the detector the best chance.
It measures the photo, not the person
Remember the analysis sees one frozen moment: this photo, this lighting, this angle. It can't see you in motion, in person, or on a good day. Two honest photos of the same face, taken minutes apart, can score differently. That's a limit of all photo-based analysis, not a bug.
And attractiveness is subjective
No automated score fully captures how attractive a face is — it's partly cultural, partly personal, and heavily influenced by things a still photo can't show, like expression, voice and presence. Use the breakdown as a guide to what you can improve, not as a verdict on your worth.
A checklist before you trust a number
- 1
Check the angle
Was your head level and facing straight at the camera at eye height? A tilt or turn moves the thirds, canthal tilt, jaw angles and symmetry.
- 2
Check the distance and lens
A close selfie distorts almost everything. The photo should be taken from about 1.5 metres with a normal lens, not a wide-angle front camera.
- 3
Check your expression
A smile, squint, pout or raised brows change the mouth, eye and brow ratios. Use a relaxed, closed, neutral face.
- 4
Check hair, glasses and light
Hair across the forehead or ears, glasses and harsh or one-sided light all interfere. Pull hair back, remove glasses and face a window.
- 5
Retake and compare
Take a second photo under the same careful conditions. If the two agree, trust them. If they disagree, one of them still has a problem.
Which measurements move most with the photo
Some rows are much more sensitive than others, so a surprising value is easier to diagnose if you know what drives it:
| Measurement | Most sensitive to |
|---|---|
| Facial thirds | Head pitch and the hairline |
| Mouth ratios and mouth corner position | Expression |
| Canthal tilt | Head roll and squinting |
| Face width to height | Lens and camera distance |
| Skin and under-eye items | Light and the time of day |
| Definition items (jaw, cheeks, chin) | Light, posture and head angle |
Questions about odd results
If two photos disagree, which do I trust?
The one taken under the recommended conditions. If both are good and still disagree on one row, that row is probably near the edge of a range, where a small change flips the score.
How can I tell whether it's my face or the app?
From one photo you can't. If a result stays the same across several good, level, evenly lit photos, it is probably real. If it jumps around, the photo is the cause.
Why does a tiny change move a score so much?
When a measurement sits close to the edge of its ideal range, a very small difference in the photo tips it across. The row's value matters more than the score for telling you what really changed.
Could the landmarks be wrong?
Occasionally, on dark, blurry or sharply angled photos. A clear, well-lit, front-on photo gives the detector the best chance of placing points accurately.
See it on a full report
Open the sample report to see every measurement, score and highlight on a real example.