First benchmark tests whether AI can actually do professional graphic design work — it mostly can't yet

Researchers built the first test suite to measure how well AI handles real design work: layouts, typography, vector graphics, animations. Current AI models pass basic understanding tasks but fail when precision matters — they can't reason about spatial relationships, generate accurate code, or handle fine typographic details.

For years, AI-for-design startups have claimed their tools can replace designers. This benchmark reveals the gap: AI understands design at a conceptual level but breaks down on the mechanical work that actually matters in production — the precise spatial math, the exact text rendering, the structured code that makes a design reproducible. It's the difference between understanding what a design should do and being able to execute it. This means current AI is a sketch tool, not a designer replacement, which matters for every company betting on automation here.