Briefle words · Prompt craft
LLM-as-a-judge
Using an AI model to evaluate or compare outputs, which works far better with explicit criteria.
Say it to an AI
Score these three app store taglines 1–5 on clarity (says what the app does), length (under 8 words) and no puns. Show a score table, total it, then pick one winner in one sentence. Don't hedge.
How an expert would ask for the stronger version of “Tagline shortlist”.
See it
Try it as a puzzle →A founder asks an AI to pick the best of three app store taglines for her dog-walking app.
Why: Without criteria A's reply judges on feel ('short, confident and modern') and marks #3 down for being clear; with them B's reply returns a scored table and a different pick.
Related words
Room 4 · Examples & references
Learn “LLM-as-a-judge” in a five-minute lesson