Across 149 completed SoftMaxx AI face scans (as of August 17, 2026), the mean overall rating is 5.4 out of 10 and the median is 4.8. Half of all scans score between 3.5 and 5.0. No scan has scored below 2.5 or above 9.0.
Those numbers surprise most people in two directions at once. The typical result sits just under the scale midpoint — not the 7/10 most people privately expect — and yet the celebrated “10/10” does not appear in the data at all. Both facts follow from how calibrated scoring works, and both are worth understanding before you read your own number.
1. The headline numbers
These statistics cover every completed scan on softmaxx.io, updated as of August 17, 2026:
- Completed scans: 149
- Mean overall score: 5.4 / 10
- Median overall score: 4.8 / 10
- Most common score band: 4.0 (22% of all scans)
- Half of all scans fall between: 3.5 and 5.0
- Share scoring 7.0 or higher: 23% (35 of 149)
- Share scoring 8.0 or higher: 14% (21 of 149)
- Lowest score recorded: 2.5 · Highest: 9.0
The mean sitting above the median (5.4 vs 4.8) tells you the distribution is right-skewed: most scans cluster just below 5, and a smaller tail of high scores pulls the average up. If you scored a 5, you are not “below average” in any meaningful sense — you are near the exact middle of real results.
2. The full score distribution
Scores are reported in half-point bands. This is the complete distribution — every completed scan, nothing sampled, nothing trimmed:
| Score band | Scans | Share |
|---|---|---|
| 2.5 | 10 | 6.7% |
| 3.0 | 4 | 2.7% |
| 3.5 | 22 | 14.8% |
| 4.0 | 33 | 22.1% |
| 4.5 | 8 | 5.4% |
| 5.0 | 12 | 8.1% |
| 5.5 | 8 | 5.4% |
| 6.0 | 9 | 6.0% |
| 6.5 | 8 | 5.4% |
| 7.0 | 4 | 2.7% |
| 7.5 | 10 | 6.7% |
| 8.0 | 6 | 4.0% |
| 8.5 | 12 | 8.1% |
| 9.0 | 3 | 2.0% |
Two features stand out. First, the floor: no completed scan has scored below 2.5, because the scoring model measures proportions and features against population reference ranges rather than handing out free-floating insults — faces simply do not measure that far outside reference ranges. Second, the ceiling: no 9.5 or 10 has ever been recorded. A 10 would require every one of the 11 measured categories to sit at its reference ideal simultaneously, which no real face does.
3. How these scores are produced
Each overall score aggregates 11 category scores — covering proportions such as facial thirds, eye-area geometry such as canthal tilt and eye shape, jaw and midface structure, skin, and symmetry — measured from an uploaded front photo (plus an optional side profile) against published reference ranges, adjusted for the measurement ranges the user selects. The full category list, reference sources and calibration approach are documented on our methodology page.
Scores are calibrated so that the scale's middle actually means something: a 5 is a face whose measurements sit near the middle of reference ranges, not a failing grade. Apps that hand every user a 7 or 8 are easier to share and harder to trust; a distribution centred near 5 is what honest calibration looks like.
4. What research says about attractiveness ratings
Two findings from the academic literature are worth knowing when reading any face-rating number:
Raters largely agree with each other. A meta-analysis of attractiveness research (Langlois et al., 2000, Psychological Bulletin) found strong agreement between raters judging the same faces, both within and across cultures — attractiveness judgments are not idiosyncratic taste, which is why measurement-based scoring is possible at all.
Judgments form almost instantly. Willis and Todorov (2006, Psychological Science) showed that attractiveness judgments made after a 100-millisecond exposure to a face correlate strongly with judgments made without time limits. First impressions of a face are fast, stable — and, importantly, they are impressions of a photo of you, which lighting, angle and expression all move.
Neither finding says a rating determines life outcomes. The measured associations between attractiveness and outcomes like wages are real but modest in the literature, and they never separate cleanly from grooming, confidence and presentation — the parts you control. See our honest guide to “am I attractive” for that discussion in full.
5. What these statistics cannot tell you
Read these numbers with their limits attached:
- Self-selection. This is not a population sample. People who scan their own face on a looksmaxxing site skew young, male, and more appearance-invested than average. A population-representative mean could sit elsewhere.
- Sample size. n = 149 completed scans. Percentages carry meaningful uncertainty at this size, and the figures will move as the sample grows. The as-of date above is part of the data.
- Mostly front-photo scans. Most scans to date are front-photo only; side-profile metrics (jaw projection, gonial angle) are absent from those results.
- A photo is not a face. Lighting, angle, lens distortion and expression all move measurements. The same face scans differently across photos — which is why single-scan scores deserve less weight than trends.
- A score is not a verdict on your life. Ratings measure geometry against reference ranges. They do not measure style, presence, fitness, grooming trajectory or anything else that changes how you actually read in person.
6. Common questions
What is the average face rating out of 10?
On SoftMaxx AI, across 149 completed scans as of August 2026, the mean overall rating is 5.4/10 and the median is 4.8/10. Half of all scans score between 3.5 and 5.0. The sample is self-selected (people who chose to scan their own face), so a population-wide average could differ.
Is a 7 out of 10 face rating good?
Yes — on our calibrated scale, only about 23% of completed scans score 7.0 or higher. A 7 sits well into the upper quarter of real results, not the middle that folk usage of “7/10” implies.
Why is the median (4.8) lower than the average (5.4)?
The distribution is right-skewed: most scans cluster just below 5, while a smaller tail of high scores pulls the mean upward. The median — the middle scan — is the better “typical result” figure.
Has anyone scored a 10?
No. The highest recorded overall score is 9.0. A 10 would require all 11 measured categories to sit at their reference ideals simultaneously, which no real face does — on an honestly calibrated scale, a perfect score should be effectively unreachable.
Do AI face ratings agree with human ratings?
Human raters agree strongly with each other about facial attractiveness (Langlois et al., 2000 meta-analysis), which is what makes measurement-based scoring possible. A calibrated AI score approximates that consensus by measuring the same proportions raters respond to — but any single scan is still a measurement of one photo, not a verdict on a face.