Prompt doctor: diagnose and fix a disappointing result
Paste an AI output that missed the mark and let ChatGPT diagnose exactly WHY — then get an upgraded prompt that fixes the root cause, not just the symptoms.
Paste this when a prompt gives you a mediocre result and you are not sure why: 'You are a prompt engineering diagnostician. I will give you a prompt I used and the result I got. Your job: 1. DIAGNOSIS — Identify the specific structural weaknesses in my prompt. For each weakness, name the failure pattern: • Missing role → generic voice • No constraints → rambling output • Vague task → surface-level answer • No format spec → unpredictable structure • Wrong abstraction level → too broad or too narrow • Missing context → hallucinated assumptions 2. ROOT CAUSE — In one sentence, tell me the single biggest reason the output disappointed me. 3. FIXED PROMPT — Rewrite my prompt with the weaknesses fixed. Use [BRACKETS] for anything I should customize. Bold the specific phrases you added and explain why each one matters. 4. BEFORE/AFTER PREVIEW — Show me a 2-sentence preview of what the fixed prompt would produce differently. My original prompt: [PASTE YOUR PROMPT] The result I got (or what was wrong with it): [PASTE THE OUTPUT OR DESCRIBE THE PROBLEM]' This is iterative prompt engineering — instead of guessing what to change, you get a structural diagnosis. Use it 2-3 times on the same prompt and it compounds. The best prompts are always version 3+, never version 1.
- Source
- promptfork seed
- License
- CC-BY-4.0
- Published
- 6/22/2026