Anatomy of an effective prompt

Lesson 3 of 8 · 30 min · part of the AI Prompt Engineer certification

Most bad outputs come from under-specified prompts, not weak models. The model fills every gap with the most generic plausible guess. Your job is to remove the guesswork.

Strong prompts are built from a handful of components:

• Role — who the model should act as (sets vocabulary, depth and perspective).
• Task — the specific action and deliverable. Use a clear verb: write, classify, extract, compare.
• Context — the facts, audience and purpose the model cannot know on its own.
• Output format — structure, length, sections, JSON schema.
• Constraints & tone — what to avoid, style, and what to do when information is missing.

Interactive exercise: Vague or specific?