They are certainly LLM-generated. I'm sure it's been done, but I would love to see a comprehensive look into what it is exactly that makes a UI seem LLM-generated.

From what I can tell, the issues are: Fonts are too small, secondary fonts have terrible contrast with their backgrounds, information is too dense/not enough white space, overuse of blinking/glowing status dots, overuse of bordered elements, and overuse of accented elements. I am surely forgetting some other issues.

I'm not sure why there seems to be a consistent set of design mistakes that LLMs tend toward.

Lack of visual hierarchy. Lots of use of bold text to make numbers stand out, without bolding the corresponding labels. Reliance on heavily-saturated primary colors. Limited attention paid to alignment.

LLMs in general are pretty terrible at spatial arrangement. You should see some of the architects' critiques of AI-generated floorplans that are on YouTube; the LLM generates a layout that is just nuts. I've found them largely useless at putting together UI mocks; Figma or Photoshop is still way easier to work with if you just want to cut up UI elements and rearrange them.

I'm not sure I would consider it too dense. If anything I think they tend to be too sparse, although this could just be that they're often used for marketing type sites, which seem to want to make you scroll and click a lot for anything anyway, even if you have the screen space.

The status dots are a part of animations in general, which they overuse, especially things like transform/scale on hover and hover effects on many many elements.

I think it's not simply borders but contrasting-color borders often with box shadows.

> I'm not sure why there seems to be a consistent set of design mistakes that LLMs tend toward.

Probably something to do with the training data and how it’s weighted. It’s the same how the writing is essentially full of overused phrases, the AI slop feel.