Most marketers now use AI, but few use it well. The gap is not the model — it is the prompt. A vague request like "write a landing page" produces generic filler; a structured prompt with role, constraints, and a success metric produces something you can ship. Below are 12 prompts I use across a full campaign cycle, grouped from basic to expert, each with a real example and the metric it affects.
Why structured prompts beat "write me a post"
The core idea is role + context + constraints + output format. Models follow instructions better when you specify the audience, the tone, the length, and what to avoid. For background on prompt patterns, see OpenAI's official prompt engineering guide (platform.openai.com/docs/guides/prompt-engineering) and Anthropic's prompt engineering documentation (docs.anthropic.com).
| Level | Prompt | Metric affected |
|---|---|---|
| Basic | Audience research | ICP accuracy |
| Basic | Positioning statement | Message-market fit |
| Basic | Headline variants | CTR |
| Basic | Product description | Conversion rate |
| Advanced | Email sequence | Open & reply rate |
| Advanced | Ad copy matrix | CPA |
| Advanced | SEO brief | Organic traffic |
| Advanced | Case study | Sales cycle |
| Expert | A/B test plan | Lift significance |
| Expert | Churn analysis | Retention |
| Expert | Competitive teardown | Win rate |
| Expert | Full funnel audit | ROAS |
Basic prompts
1. Audience research
Task: turn raw customer data into a usable ICP.
Prompt: "Act as a B2B market researcher. Here are 15 customer interview notes: [paste]. Identify 3 segments, each with: primary pain, buying trigger, objection, and one verbatim quote. Output as a table. Do not invent quotes."
Example result: three segments — agencies, in-house teams, solo founders — with distinct triggers. Use it to brief writers; it typically cuts research time from a day to under an hour.
2. Positioning statement
Task: convert features into a positioning line.
Prompt: "Write 3 positioning statements for [product]. Format: For [audience] who [need], [product] is a [category] that [benefit], unlike [alternative]. Keep each under 25 words."
3. Headline variants
Task: generate testable headlines.
Prompt: "Generate 10 headlines for [page]. Mix benefit-led, curiosity-led, and number-led. Max 12 words each. Flag which psychological trigger each uses."
4. Product description
Prompt: "Write a 120-word product description for [item]. Tone: [tone]. Include one concrete use case and one objection handled. No superlatives like 'revolutionary'."
Advanced prompts
5. Email sequence
Task: build a 5-email onboarding chain.
Prompt: "Create a 5-email onboarding sequence for [product]. For each: subject line, preview text, body under 150 words, single CTA, and send day. Goal: activation, not upsell."
6. Ad copy matrix
Prompt: "Generate 12 ad variants for [offer] across 3 angles (pain, proof, urgency) and 4 lengths. Output CSV: angle, headline, primary text, CTA."
7. SEO brief
Prompt: "Create an SEO brief for the keyword '[keyword]'. Include search intent, 5 H2s, entities to mention, related questions, and internal link targets. Base it on the current top 5 results, not assumptions."
8. Case study
Prompt: "Turn these notes into a 500-word case study: [notes]. Structure: context, problem, approach, result with numbers, quote. Keep claims only to what the notes support."
Expert prompts
9. A/B test plan
Prompt: "Design an A/B test for [page]. Specify hypothesis, primary metric, sample size logic, duration, and what would falsify it. Warn me if expected lift is below detectable effect."
10. Churn analysis
Prompt: "Analyze these churn reasons: [list]. Group into themes, rank by frequency, and propose one retention experiment per top theme."
11. Competitive teardown
Prompt: "Compare [competitor] and [us] across pricing, positioning, onboarding, and proof. Output a table plus 3 gaps we can exploit."
12. Full funnel audit
Prompt: "Audit this funnel: [steps and metrics]. Identify the biggest drop-off, estimate revenue impact, and suggest 3 fixes ranked by effort."
How to use this without losing your voice
AI drafts, humans decide. Always verify numbers, quotes, and claims before publishing. Treat every output as a first draft that needs your judgment — that is what keeps quality high while speed improves.
If you want to go deeper, practice these prompts on real campaigns and measure the result. The marketers who win in 2026 are not the ones with the best model — they are the ones with the best questions.
Comments