People asking for a brutally honest face rating are usually reacting to something specific: every rating they have had so far felt like it was designed to make them feel good. Friends round up. Dating apps have no reason to tell you anything. Plenty of rating tools return a 7.5 to almost everyone, because a satisfied user is a paying user.
So it is worth being precise about what honest means here, because it is not the same as harsh. Honest means calibrated: the scores a tool hands out should have the same shape as the population it is scoring. And a calibrated rater turns out to be less brutal than people expect, in both directions.
What a calibrated distribution looks like
Across 17,032 analyses run between January and August 2026, the mean score was 5.59 and the median was 5.3.
| Threshold | Share scoring at or above |
|---|---|
| 5 | 74.7% |
| 6 | 25.1% |
| 7 | 8.1% |
| 8 | 3.3% |
| 9 | 0.2% |
Broken into bands, 4,230 people scored between 4.0 and 4.9, 8,450 between 5.0 and 5.9, 2,891 between 6.0 and 6.9, 824 between 7.0 and 7.9, 524 between 8.0 and 8.9, and 36 at 9.0 or above. Just 77 people, 0.5% of the dataset, scored below 4.0.
That is the shape honesty has to match. Most people near the middle, thin tails on both sides.
Why a 7.5 for everyone is flattery
Run those numbers against a tool that hands out 7.5s. In this dataset, the entire 7.0 to 7.9 band contains 4.8% of people, and only 8.1% clear 7.0 at all. A rater that puts the typical user at 7.5 is placing the average person in a band that roughly 1 in 20 people occupy.
When that happens, the number has stopped carrying information. If nearly everyone gets between 7 and 8.5, the score no longer distinguishes anyone from anyone, and the only thing it communicates is that the tool would like you to come back. Inflation is not generosity. It is the removal of the one thing you asked for.
The same logic applies to social ratings. A friend asked to rate your photo is answering a social question, not a measurement one, and the socially correct answer is high. That is not dishonesty on their part. It is just a different question.
Brutal is not the same as accurate either
The opposite failure is less common but worth naming. A rater that mostly returns 3s and 4s to look tough is miscalibrated in the other direction and equally uninformative.
The data does not support that reading of people. Only 0.5% of 17,032 analyses fell below 4.0. The bottom of the scale is nearly as empty as the top. Anyone whose honest expectation of themselves is a 3 is, on this evidence, almost certainly wrong, and a tool that confirms it is not being rigorous.
This is the part that gets lost in the phrase "brutally honest". The same calibration that refuses to hand out 8s also refuses to hand out 3s. Those two refusals come from the same place.
What honest feedback actually points at
A calibrated score is only half of it. The more useful half is what the number decomposes into, and the pattern there is consistent.
Symmetry correlates with the overall rating at 0.95, the strongest relationship in the data. People scoring 8 or above overall average 8.54 on symmetry; people under 5 average 4.74. Golden-ratio compliance, which dominates looksmaxxing content, averages 77.28% across the dataset with a maximum of 96%, and correlates with the overall score at only 0.67.
The reading of that is fairly blunt. Chasing an exact classical proportion is chasing the weaker signal, and nobody in 17,032 faces reaches the ratio anyway. Left-right consistency and general facial harmony move the number far more, and both are heavily affected by the photograph itself - angle, camera height, lighting and expression all show up as asymmetry in the score.
Which means the first honest thing to check is not your face. It is whether the picture is a fair record of it.
Getting a rating you can trust
Shoot front-on at eye level, in even light, with a neutral expression and hair off the face, and take more than one. If three photos return three very different scores, the variation is telling you about the photography rather than about you.
Then read the result against the table above rather than against 10. A 5.3 is the median. A 6.0 is in the top quarter. A 7.0 is roughly 1 person in 12. Anything at 8 is top 3.3%.
Two tools do this with the distribution attached. The attractiveness test returns the score with your percentile against these 17,032 analyses, so the placing is done for you rather than left to guesswork. The ChatGPT face rating page covers what happens when you ask a general chat model instead, and why a conversational assistant tends to soften a number that a calibrated tool does not.
The honest conclusion from all of this is not a compliment and not an insult. Half of everyone in this dataset scored between 5.0 and 5.9. A score in the fives is the single most common outcome there is, which is what "average" means when it is measured rather than felt. The full percentile breakdown is in the average face rating study.