Avoiding Common Mistakes
The 7 most common prompt mistakes
1. Too vague — Help me with my marketing. Fix: add task + audience + format.
2. Too long — Walls of text confuse the model. Fix: one task per prompt.
3. Assuming shared context — Continue from before. The AI has no memory across sessions. Fix: always re-state the context.
4. No format instructions — leads to random structure. Fix: always specify format.
5. Negative instructions only — Don't be too formal. Fix: add what you do want: Use a friendly, conversational tone.
6. Multiple conflicting goals — Write a short but comprehensive guide. Fix: pick one or prioritize.
7. Not iterating — accepting the first output without refining. Fix: always do one follow-up.
The power of iteration
The best prompt engineers rarely get perfect output on the first try. They iterate:
1. Send prompt → get output
2. Identify what's wrong or missing
3. Add a correction or follow-up prompt
4. Repeat until satisfied
Common follow-ups:
• Make it shorter / longer
• More formal / casual
• Add an example for point 3
• Rewrite the intro — it's too generic
• Translate this to Arabic and keep the same tone
Key Takeaways
- Vague, long, or conflicting prompts produce poor output.
- Always re-state context — AI has no memory between sessions.
- Specify what you want, not just what you don't want.
- Iteration is normal — the best output comes after 2–3 refinements.
Fix a broken prompt
Here's a bad prompt. Identify at least 3 mistakes and rewrite it: `Write content for my social media about my shop that sells things. Make it good and professional but also fun and short but detailed.`