Write a marketing plan.
No role. No context. No format. The model has to guess what kind of "plan" you want — so it picks the most generic one.
Act as a {{senior B2B marketing strategist}}. # Context - Product: {{$49 prompt pack for freelancers}} - Audience: {{solo freelancers, 6 mo–4 yr in}} - Budget: $1,500/mo - Channels available: LinkedIn, X, newsletter # Task Build a 90-day acquisition plan in phases: 1. Audience clarity 2. Offer + messaging 3. Channel sequencing 4. Weekly content cadence 5. Conversion path # Format Phase header → key actions → metric. Flag risks at each phase.
Role, context, task, format, constraints. The model has everything it needs to give a sharp answer.
Prompt types help you choose the right level of instruction. Sometimes you need one clean answer. Sometimes a strategy. Sometimes a complete workflow. Sometimes multiple specialist agents working together.
Learn the formula →A direct, clearly formatted prompt designed to produce one specific output. Use it when you already know what you need and want the model to produce it quickly.
# Role Act as a {{[ROLE]}}. # Task I need help with {{[TASK]}}. # Context {{[BACKGROUND INFORMATION]}} # Output Create {{[SPECIFIC OUTPUT]}}. Format it as {{[FORMAT]}}. # Constraints {{[CONSTRAINTS]}}
Act as a {{professional email writer}}. Write a follow-up email for a meeting I attended. # Context - Meeting topic: {{topic}} - Attendees: {{list}} - Key decisions: {{decisions}} - Action items: {{owners + deadlines}} - Open questions: {{open items}} # Format Subject · Opening · Decisions · Actions Open items · Clear next step. # Constraints - Professional, concise, scannable. - No corporate jargon.
Tells the AI to act as a specialist — advisor, strategist, auditor, coach, expert. Instead of just producing output, the model thinks from a professional role. Useful when the task needs judgment, prioritization, critique, or expert reasoning.
# Role Act as a {{[SPECIFIC EXPERT ROLE]}}. # Goal Help me {{[OBJECTIVE]}}. # Context - {{[DETAIL 1]}} - {{[DETAIL 2]}} # Reasoning chain 1. Identify the main issue. 2. Explain the root cause. 3. Recommend the best option. 4. Give an action plan. 5. Flag risks or assumptions. # Output Diagnosis · Recommendation Action plan · Risks · Next steps
Act as a {{senior SEO content strategist}}. # Situation - Niche: {{niche}} - Audience: {{audience}} - Problem: {{traffic flat, blog not converting}} # Analyze in 5 steps 1. Identify the likely SEO problem. 2. Explain the cause. 3. Suggest top 5 content opportunities. 4. Recommend which pages to update first. 5. Build a 30-day action plan. # Output Diagnosis · Content gaps Priority pages · 30-day roadmap Mistakes to avoid.
Breaks a larger task into a sequence of steps. Instead of asking the AI to do everything at once, you guide it through phases: research → plan → draft → improve → finalize.
# Complete this task in phases STEP 1 — {{[FIRST PHASE]}} Analyze {{[INPUT]}} and produce {{[OUTPUT]}}. STEP 2 — {{[SECOND PHASE]}} Use Step 1 output to create {{[OUTPUT]}}. STEP 3 — {{[THIRD PHASE]}} Use Step 2 output to create {{[OUTPUT]}}. STEP 4 — {{[FINAL PHASE]}} Review everything and produce the final deliverable. # Rules - Do not skip steps. - Ask clarifying questions if needed. - Always feed the previous output forward.
Act as a senior content strategist. Brand: {{brand}} Audience: {{audience}} Platform: {{platform}} STEP 1 — Audience + goal Analyze audience, buying intent, content goal, key pain points. STEP 2 — Content pillars Build 4 pillars with angle and conversion path for each. STEP 3 — Calendar 30-day calendar: Day · Topic · Format · Hook · CTA · Pillar. STEP 4 — Repurpose Pick top 5 highest-potential posts and suggest short-form, carousel, and email versions.
Multiple specialist AI roles in sequence — each agent handles one part of a larger workflow. A blog production system might include a Keyword Researcher, Outline Architect, Content Writer, and Editor/SEO Optimizer.
# 4-agent workflow INPUTS - Brand: {{brand}} - Audience: {{audience}} - Goal: {{goal}} AGENT 1 — {{[ROLE]}} Input: {{[INPUT]}} Task: {{[TASK]}} Output: {{[OUTPUT]}} AGENT 2 — {{[ROLE]}} Input: Agent 1 output Task: {{[TASK]}} AGENT 3 — {{[ROLE]}} Input: Agent 2 output # Rules - Run in sequence. - Each agent uses the previous output. - Do not skip agents.
AGENT 1 — Keyword Researcher Task: Top 5 long-tail keywords, intent, reader problem, best angle. Output: Keyword list + 3 title options. AGENT 2 — Outline Architect Input: Use Agent 1 output. Task: Full outline with H1, intro hook, 5-7 H2 sections, CTA, meta description. AGENT 3 — Content Writer Input: Use Agent 2 output. Task: Write the post following outline. AGENT 4 — Editor & SEO Input: Use Agent 3 output. Task: Score, fix weak sections, internal links, readability pass.
If you're new to AI prompting, learn them in this order. Don't start with complex multi-agent systems if you're still figuring out the basics.
Before any prompt type, learn to define Role · Context · Task · Format · Constraints. The 5-part formula is the foundation under everything below.
Learn the formulaThe four main types are Structured, Agentic, Multistep, and Multi-Agent. Structured prompts are best for fast outputs. Agentic prompts are best for expert thinking. Multistep prompts are best for workflows. Multi-agent prompts are best for advanced systems.