I’ve always felt that traditional "Which character are you" quizzes were too static—relying on rigid branching logic and simple scoring that often yields predictable, low-accuracy results.
To solve this, I built CharacterTest.app. Instead of simple keyword matching, it uses an LLM-powered engine to perform semantic personality mapping. We decode user responses into the Big Five (OCEAN) and MBTI dimensions, then calculate proximity within a high-dimensional trait database covering 2,000+ characters across 100+ universes.
Technical Stack:
Framework: Next.js (App Router) with SSR for sub-second page loads and SEO.
AI Logic: Custom prompting architecture to prevent "personality bleed" and ensure multi-language consistency (supporting EN, JP, KR).
Privacy: No signup required for basic tests. I believe in "value first, data later."
I’m particularly curious about your thoughts on the mapping accuracy and the latency of the AI matching process. Would love to hear your feedback!