A 5 out of 10 is average - genuinely, measurably average. Across 17,032 AI face analyses the mean score is 5.59 and the median is 5.3, and the 5.0 to 5.9 band alone contains 8,450 people, or 49.6% of everyone analysed. Very nearly half of all results land in that one band.
If you scored a 5, the most accurate thing that can be said is that you got the most likely outcome there is.
Is a 5 below average or exactly average?
Very slightly below the middle. 74.7% of people score 5.0 or above, so roughly three quarters of results sit at or over a 5, and the median is 5.3 - meaning a flat 5.0 is a fraction under the halfway point of real scores.
The distinction barely matters in practice. What matters is that the mental model most people bring to this is wrong. They assume 10 is the scale and therefore 5 is the midpoint, so a 5 reads as "half of everyone is better looking than me". On a scale calibrated so that real faces cluster around the middle, a 5 is simply the most ordinary result there is. If the worry underneath the question is the harsher one, the am I ugly test takes it on directly: only 0.5% of people score below 4.
Why do half of all people score a 5?
Because the distribution is tight, not flat. Scores do not spread evenly from 1 to 10 - they bunch heavily in the middle and fall away fast at both ends.
| Score band | People | Share |
|---|---|---|
| Below 4.0 | 77 | 0.5% |
| 4.0 - 4.9 | 4,230 | 24.8% |
| 5.0 - 5.9 | 8,450 | 49.6% |
| 6.0 - 6.9 | 2,891 | 17.0% |
| 7.0 - 7.9 | 824 | 4.8% |
| 8.0 - 8.9 | 524 | 3.1% |
| 9.0 and above | 36 | 0.2% |
Three quarters of everyone sits between 4.0 and 5.9. Only 0.5% score below 4.0 and only 3.3% score 8.0 or above.
The scale is doing what a well-calibrated scale should: reserving the extremes for genuine extremes. A tool that hands out 8s freely is not measuring anything, it is flattering you. That is worth remembering when a different site tells you the average is 7.4 - that figure describes their scale, not their users.
What a typical 5 actually looks like
The 5 band is close to the definition of a statistically ordinary face, and several independent measures agree on that.
Golden-ratio compliance for the 5.0-5.9 band averages 77.0%, against a population average of 77.3%. Facial symmetry across the whole dataset averages 5.62. And the two dominant face shapes - oblong at 44.4% and oval at 42.3%, together 86.7% of everyone analysed - average 5.55 and 5.66 respectively, both within 0.11 of the overall mean.
There is one more finding worth knowing here, because it reframes what improvement means. The average gap between a person's best and worst feature is only 1.38 points, and 58% of people tie for their weakest feature rather than having a single obvious one. Most people do not have one catastrophic flaw dragging a good face down. They have twelve features clustered closely together, slightly below where they could be.
How much can a 5 improve?
The 5.0-5.9 band averages 1.36 points of headroom with an average potential score of 6.66. That is a realistic move from the middle of the pack into the top quarter, since only 25.1% of people score 6 or above.
It is also close to the population-wide average gap of 1.31 points, which is the honest ceiling for almost everybody. The full spread of available headroom looks like this.
| Headroom available | Share of people |
|---|---|
| Under 0.5 points | 2.8% |
| 0.5 - 1.0 points | 14.4% |
| 1.0 - 1.5 points | 36.7% |
| 1.5 - 2.0 points | 41.7% |
| 2.0 points or more | 4.4% |
The striking part is how narrow that is. The standard deviation of the gap is just 0.42 points, so the amount of improvement available is remarkably consistent from person to person. Whatever your starting point, the model rarely thinks you have less than a point available, and it almost never thinks you have more than two.
Nobody in the dataset gained 4 points, and only 15 people out of 17,032 had a gap of even 3.
What to actually do with a 5
The same dataset points at the answer, and it is unglamorous. Skin is the weakest feature for 38% of people - two and a half times more common than the next-most-frequent weak point - and the lowest-scoring of the twelve features on average at 5.30. It is also fully improvable, which bone structure is not.
Beyond that, the largest single-day swing available to most people is the photograph. Even lighting at eye level, a neutral background and a natural expression are worth real points, and they cost nothing.
The complete percentile table is in the average face rating study, the headroom data is in the potential vs actual study, and the face-shape figures come from the face shape study. To see where you sit, the face rating tool returns your current and potential scores together, and if you scored just above this band, is a 6 out of 10 a good face rating covers the next step up.