A simple prompt iteration checklist

When an answer misses the mark, change one variable at a time. A short diagnosis makes the next prompt both smaller and more effective.

Check the goal

Could another person tell what the output is meant to accomplish? Clarify the decision, action or reader response you want.

Check the inputs

Look for missing audience context, source material, constraints or examples. Do not ask the model to guess information you already have.

Check the failure precisely

Replace “this is bad” with an observable critique: the examples repeat the same idea, the tone is too formal, or the recommendation ignores the budget.

Ask for a targeted revision

Keep the useful parts and revise only the failure. This preserves momentum and shows which instruction actually improved the result.