01 , Case Study
[ LAB ]
AI · SAAS · WEB APP · 2026
Most website feedback is either too polite to be useful, too expensive to be accessible, or too generic to be actionable.
TIMELINE
3 weeks
ROLE
Design + Engineering
STACK
Next.js, vision models, LLM analysis
OUTCOME
Live SaaS, public tool
02 , The Problem
課題Most website feedback is useless. Friends are polite. Agencies sugarcoat to upsell. Generic audit tools spit out meaningless scores. Meanwhile, the average visitor to a website makes a judgment in three seconds — and then leaves without telling you why.
We wanted to build the opposite: a tool that would look at a website and tell you exactly what was wrong with it, plainly, with no hedging.
03 , Approach
手法Roast My Website feeds a screenshot of any public site into a vision-enabled LLM, which generates a structured critique across design, copy, usability, and positioning. The output is deliberately blunt. No scores, no percentage grades, just direct feedback written the way a senior designer would deliver it to a peer.
“The most valuable feedback is the feedback nobody wants to hear.”
Gallery
04 , Outcome
成果Roast My Website is a live tool, free to try. It's been used on thousands of sites.
NEXT