Some details on how you did this would be nice.

Looking at, say, the Middle East it's crazy it's giving almost the same dress and garb for multiple countries - same for some African countries.

I would imagine that if I simply pick one country and ask it repeatedly to make the stereotypical man/woman, it would give me quite different appearances/clothes each time. So this is very suspicious (or a bug).

I asked each model a simple prompt: "Most stereotypical woman in XXX. No text. " I didn't edit anything, just one attempt. I'd expect some variety too, but it does show some variety/biases in the model's image generation~

The source code is available here: https://github.com/anvaka/imaginary-faces

Very interesting. Hi from Argentina!

Midjourney: (Is that a cat?) All of them look very poor and dirty. I don't want to claim everybody here is rich, but it looks like the images after a disaster and I'm not sure they are very stereotypical.

Grok: She is cute, but I'm not sure she is from Argentina. He looks right, but the style of carrying the vacuum flask is a little more Uruguayan.

GPT: She looks like an influencer, a little meme heavy but plausible. He looks like a collage of a 5 years old for school where the kid has to paste every single national stereotype, I'll add a small picture of Kempes+Maradona+Messi just to go to 12 level.

Gemini: In both case the idea and composition is right, but there is something weird with the texture, like if they were photos taken with a 1990 camera instead of a 2020 phone.

I all cases, they look like photos taken in Buenos Aires (city) and the gaucho looks like a gaucho of Buenos Aires (province). It's the biggest city, so it's probably fine for "stereotypical".

PS: You have no favicon.