Brandolini’s Law — the bullshit asymmetry principle — states that refuting nonsense requires an order of magnitude more effort than producing it. That ratio was already brutal. AI collapsed the production cost to near zero, and the consequences are hitting design hard.
The problem isn’t that AI generates bad work. It’s that AI generates plausible-looking bad work at volume, and the effort to distinguish it from good work stays constant — or increases. For designers, this shows up in specific ways:
- AI-generated UI mockups flood stakeholder reviews with options that look polished but ignore interaction logic, accessibility, and edge cases
- Auto-generated copy and microcopy reads fine in a screenshot but falls apart in real user flows — error states, empty states, multi-step forms
- Design system contributions from AI tools create components that visually match a system’s tokens but break semantic structure, naming conventions, or responsive behavior
- Portfolio and hiring noise — recruiters now sift through AI-assembled case studies that mimic the shape of design thinking without containing any
The asymmetry compounds. A junior designer can generate 40 screens in an afternoon with Figma AI or v0. A senior designer still needs the same time to audit whether those screens actually solve the user’s problem. The bottleneck moved from production to judgment, but organizations haven’t adjusted their processes or headcount to match.
This isn’t a quality-filter problem you can automate away. The whole point of Brandolini’s Law is that detection is inherently harder than generation. Adding an AI layer to catch AI slop just adds another source of plausible-looking output that needs human review.
NOTE
The original essay draws a distinction between “slop” (low-effort AI output passed off as real work) and legitimate AI-assisted design. The issue isn’t using AI tools — it’s the economic incentive to skip the judgment step when production is free.
The practical pressure lands on design leads and hiring managers. Review processes built for a world where production was expensive — where volume itself signaled effort — need restructuring. Critique sessions need to evaluate reasoning, not just output. And teams need explicit policies on when AI-generated artifacts require human validation before they enter a design system or ship to users.