Mastering Imperative Prompts:
A Command-Driven Guide

Learn how to control LLMs, boost compliance, and generate radical results.
Only real, research-backed techniques.

Enter a prompt and tap Enhance!

Why These Tactics Work

AI's Favorite Verbs: AIs are trained on tons of examples where words like 'List', 'Write', or 'Generate' mean 'do this specific thing.' They learn this pattern really well and know it means “get to work.”
Command First = AI Listens Better: AIs pay most attention to words at the start of your prompt. Put your main command ('List the reasons...') upfront so the AI gets the point immediately and doesn't get sidetracked.
Clear Verbs = Clear Results: 'List' tells the AI to make a list—super clear. 'Explore' is fuzzy; the AI doesn't know what an 'exploration' output should look like, so it might just ramble. Specific verbs get predictable results.
AI Training Echo Chamber: A lot of AI training data is made by *other* AIs. This means command words and sentence styles get repeated and reinforced, making them extra reliable.

Golden Prompt Formula

COMMAND + THING + FORMAT

Example: Generate 5 startup ideas in a table with name, audience, risk

Enter a prompt and tap Diagnose.

Prompt-Strengthening Techniques

  • Upgrade Soft Verbs: "Reflect on..." → "List 3 reflections and rank them."
  • Use Roles: "You are a cautious thinker. Rank 3 risks."
  • Role Prompt Hack: "You are a prompt engineer. Evaluate compliance."
  • Instruction Spotlighting: Use brackets: "Please [Generate] a summary."
  • Stepwise Refinement: Start with hard command, then refine (generate → bullet → analyze).
  • Audience Framing: Vary tone: explain to a 10-year-old vs. PhD.
  • Chained Workflow: Idea → Evaluate → Improve.

Verb Tier Table

TierBinding StrengthTypical VerbsBehaviorExampleWhy it WorksFix Guidance
S-TierHighGenerate, Create, Write, List, Summarize, Define, TabulateDirect response, predictable structureList 3 startup ideas in a tableCommon in training data, matches output expectationsUse when output must be guaranteed
B-TierMediumRefine, Edit, Adjust, Evaluate, Compare, Explain, Analyze, ClassifyWorks when you define what to edit & output formatRefine summary into bullet pointsImplies a task, but AI needs more contextAdd noun and format to clarify
D-TierLowExplore, Consider, Reflect, Brainstorm, Workshop, Discuss, SuggestAI may ramble, output is variableBrainstorm startup ideasSoft, unclear—AI guesses or ramblesRewrite with S-Tier verb, e.g., 'List 5 ideas for X'
F-TierAvoidImagine, Wonder, Muse, Speculate, Dream, FeelUnpredictable, creative fluff or nonsenseImagine the future of AINo clear pattern in training dataOnly for fun or creative use

Adverb Control Grid

Verb+ Briefly+ Logically+ Clearly+ Formally
ExplainIn 2-3 linesStep-by-step reasoningNo jargonPolished tone
Summarize≤ 100 wordsBullet outlinePlain EnglishExecutive format
Describe1 paragraphOrdered by featuresFocus on visualsNeutral, objective

Troubleshooting & Diagnostics

Rambling output: Used soft/introspective verb (like "think", "explore").
Fix: Replace with S-Tier verb + noun + format.

Model ignores "don't include X": Negative constraint poorly processed.
Fix: State positive directive, e.g., "Include only X".

Unstructured response: Missing format specifier.
Fix: Add: "as 3 bullet points", "in table format", etc.

Verbose/dry: Default temperature used.
Fix: Use 0.2 for logic, 0.8 for creativity.

Prompt Fix Library

BeforeAfter
Let's explore UI designs.Design a minimalist UI; provide 3 layout bullets.
Can you think about revenue ideas?List 5 revenue ideas and rank by ROI.
Please workshop a prompt with me.Ask 3 questions to refine my image prompt.
Consider adding suspense.Rewrite scene; insert suspense beat after paragraph 2.
Reflect on training data bias.Summarize 3 key training data biases in 1 sentence each.

Glossary

TermDefinition
TemperatureControls randomness. 0 = deterministic, 1 = creative randomness.
Few-shot PromptingProviding examples to guide output format.
Chain-of-ThoughtStep-by-step reasoning scaffold.
Prompt RigidityHow strictly the model follows format.
Hard PromptFormula: Verb + Noun + Format + Constraint.

Summary Rules

  • Use Tier-1 verbs: Generate, Write, List.
  • Follow the formula: Verb + Noun + Format + Constraint.
  • Soft verbs lead to vague results—wrap them in hard scaffolds.
  • Fix prompt failures with clear format and structure instructions.
  • Each model behaves differently: GPT-4 prefers structure, Claude adds caution, Gemini returns facts.

Command = Control.

Stop asking nicely. Start telling the AI what to do.