Does ChatGPT rate everyone an 8?

Three things push a general assistant towards the same comfortable number. It is trained to be helpful and kind, and a low score about someone's face reads as neither. It has no anchor: nothing tells it where an average face should land, so it falls back on the way people talk, where 7 out of 10 is the polite default. And it rarely sees the full range of faces in one session, so it cannot place your photo against anything.

The result, as a general tendency rather than a measured statistic, is a narrow band of flattering scores. A number that almost everyone gets carries almost no information, which is why the 8 you were given and the 8 your friend was given may describe very different faces.

The same photo can come back a 6, then a 7.5, then a refusal. Wording does a lot: "be brutally honest" pulls the number down, "rate me kindly" pulls it up, and asking for a breakdown first can change the total. A different crop, a new chat or a follow-up question can shift it too. None of that is the face changing. It is the absence of a fixed scale showing.

A useful scale states its average out loud and sticks to it. BecomeTen scores out of 100 and is calibrated so that an average, healthy, groomed adult lands between 50 and 60, which is about a 5 on the ten-point scale. Scores from 78, the High tier, are reserved for striking structure, so they are rare. The attractiveness scale shows how the 0–100 score, the tiers and the 1–10 instinct line up, the PSL scale covers the forum version, and why average is 50 explains the choice.

So if a chatbot gave you an 8 and an anchored rater gives you a 55, the two are not contradicting each other as much as it seems. One is a compliment on a scale where everyone is near the top; the other is a position on a scale with room above it. For what an honest average means, see what is average attractiveness, and for the wider picture, can ChatGPT rate my face.