Anti UI Slop

The core insight — read this before any rule

AI-generated UIs are recognizable not because of any single CSS property, but because they converge. Every generator, asked for "a modern clean UI", reaches for the same defaults: the same fonts, the same purple gradient, the same icon-tile-above-heading card, the same hero→metrics→features template. In 2022 it was purple gradients and glassmorphism; in 2024 it was cream backgrounds and Instrument Serif italic heroes. The specific tells change every year. The disease is constant: a reflex where a decision should be.

This has two consequences that govern everything below:

1. Deleting a tell is not a fix. If you remove the purple gradient and reach for the next "safe" default, you have produced next year's slop. Every fix must substitute a deliberate choice derived from this product's context — its domain, brand, audience, content. Before editing anything, articulate (in one or two sentences) what this interface's design direction actually is. Then make every fix serve that direction.
2. Slop rules are judgment calls; quality rules are not. An italic serif headline is legitimate on an editorial magazine and a tell on a SaaS landing page — judge by context. But low contrast, broken images, and tiny text are objectively broken for every user, always. The two categories below are handled differently.

A reliable smell test for your own output: if you could swap this UI onto a different product and nothing would feel wrong, no decisions were made. Distinctive design is specific — it could only belong to this product.

Workflow