You are a senior marketing strategist. Write a one-page campaign brief for the launch below.
Product: [PRODUCT NAME + ONE-LINE DESCRIPTION]
Primary audience: [WHO THEY ARE + 1 KEY PAIN POINT]
Goal: [SPECIFIC, MEASURABLE — e.g. "300 trial signups in 30 days"]
Budget + channels: [TOTAL $ AND CHANNELS YOU ALREADY USE]
Timeline: [START → END DATE]
Constraint: [LEGAL, BRAND, OR INTERNAL LIMIT]
Output structure:
1. Campaign objective (1 sentence, with the metric)
2. Core insight about the audience (2 sentences)
3. Single-minded message (under 10 words)
4. 3 proof points the message rests on
5. Channel mix with rationale for each
6. One risk and how to mitigate it
Plain language. No buzzwords. Nothing you can't measure.
Why this works
A brief is only useful if it forces decisions. This prompt makes the model commit to one objective, one message, and one risk — instead of producing a soup of ideas. The "no buzzwords, nothing you can't measure" guardrail at the end kills the AI's instinct to pad with words like synergize and holistic. Run it twice and merge the strongest answers.
Write 5 ad headline variations for the offer below, each using a different angle.
Product: [WHAT IT IS]
Audience: [WHO YOU'RE TALKING TO]
Main benefit: [WHAT CHANGES FOR THEM]
Offer: [PRICE, TRIAL, DISCOUNT, OR HOOK]
Channel: [META / GOOGLE / LINKEDIN / TIKTOK]
Word limit: [E.G. UNDER 40 CHARACTERS FOR GOOGLE]
Use these 5 angles, in order:
1. Pain-point — name the problem in their words
2. Outcome — paint the after-state in concrete terms
3. Social proof — reference numbers, customers, or results
4. Curiosity — tease something they don't know yet
5. Urgency — give a real reason to act now (no fake deadlines)
For each headline, output: the angle, the headline, and one line on who it's for. Skip clichés ("revolutionary," "game-changing," "unlock").
Why this works
Asking for "5 ad headlines" gets you 5 variations of the same idea. Forcing 5 distinct angles — pain, outcome, proof, curiosity, urgency — guarantees real diversity for testing. The cliché blocklist matters: AI loves words like
revolutionary, and disabling them upfront saves an editing pass. Pair with the
SEO prompts for keyword-aware variants.
Build a one-page brand voice guide based on the inputs below. The goal: anyone on the team (or any AI) can write in this voice after reading it.
Brand: [NAME + ONE-LINE WHAT YOU DO]
Audience: [WHO THEY ARE + WHAT THEY VALUE]
3 brands whose voice we admire: [LIST]
3 brands whose voice we'd never sound like: [LIST]
Sample of our writing we like: [PASTE 2–3 SENTENCES]
Output:
1. Voice in 3 adjectives (with a one-line "why this, not that" for each)
2. We sound like / we don't sound like — 5 paired examples
3. Punctuation + formatting rules (em dashes, exclamation points, emoji, sentence length)
4. 3 "always say" phrases and 3 "never say" phrases
5. A 60-word sample paragraph in this voice about [SAMPLE TOPIC]
Be opinionated. A voice guide that says "friendly but professional" is useless.
Why this works
"Brand voice" prompts usually fail because they ask for adjectives in a vacuum. This one anchors the model with reference brands you admire and reject — the contrast is what makes the output specific. The "we sound like / we don't" pairs are gold for onboarding writers. Save the output and paste it as the system prompt for every future writing task.
Build a 4-week content calendar for [BRAND] across [CHANNELS — e.g. blog, LinkedIn, email].
Audience: [WHO + WHAT THEY CARE ABOUT]
Business goal this quarter: [E.G. "DRIVE TRIALS FOR PRODUCT X"]
Pillars (3–5 themes we own): [LIST]
Cadence: [POSTS PER WEEK PER CHANNEL]
Voice: [LINK OR SHORT DESCRIPTION]
For each week, output a table with these columns:
- Date
- Channel
- Pillar
- Working title
- Hook / first line
- Format (essay, list, case study, thread, etc.)
- CTA
- Why this fits the goal
Mix 70% educational, 20% credibility (case studies, data), 10% promotional. No filler topics; every row should clearly tie to a pillar and the quarter goal.
Why this works
Most content calendars die because every post is a one-off idea. Forcing the model to slot every row under a pillar and tie it to the quarter goal stops the drift. The 70/20/10 mix is a useful default for B2B — adjust if you're consumer. Treat the output as a draft, not a contract: kill the rows that feel weak, then run the prompt again with sharper pillars.
Build a usable customer persona based on the inputs below. Skip generic demographics; focus on what actually changes the marketing.
Product: [WHAT IT IS]
Best 3 customers we've talked to: [SHORT NOTES — JOB, WHY THEY BOUGHT, WHAT THEY SAID]
Where they hang out online: [PLATFORMS, NEWSLETTERS, COMMUNITIES]
What they were doing before our product: [PRIOR SOLUTION OR WORKAROUND]
Output a one-page persona with:
1. Persona name + one-line "who they are at work"
2. The job they're trying to do (Jobs-to-be-Done frame)
3. The trigger event that puts them in-market
4. 3 objections they raise before buying
5. What "good" looks like to them after 30 days with us
6. 3 message angles that would land + 1 angle that would backfire
7. Where to reach them (rank 3 channels by likely cost-efficiency)
Use real verbs and nouns from the customer notes. No corporate clichés.
Why this works
Demographic personas ("Marketing Mary, 34, suburban") don't change a single decision. JTBD-style personas do, because they tell you the trigger, the alternative, and the objection. The instruction to use real verbs from customer notes is what keeps the model honest — without it, you get the corporate-handbook persona we've all read 1,000 times.
Draft a long-form landing page for the offer below, using the PAS-Solution-Proof-Action framework.
Product: [WHAT IT IS]
Target visitor: [WHO + WHERE THEY'RE COMING FROM — AD, ORGANIC, REFERRAL]
Primary goal: [SIGNUP, DEMO, PURCHASE]
One thing the visitor needs to believe to act: [BELIEF]
Output sections (in this order):
1. H1 hero (8–14 words, names the outcome)
2. Subhead (1 sentence, names the audience)
3. Problem section — 3 bullets in their voice
4. Agitation — 1 short paragraph on the cost of staying stuck
5. Solution — what we do, in 3 plain-language steps
6. Proof — 3 placeholders for testimonials, data, or logos
7. FAQ — the 5 most likely objections
8. Final CTA — the action + a one-line reason to act now
Keep paragraphs to 3 sentences max. Write at an 8th-grade reading level. No "imagine" openers.
Why this works
Landing pages fail when the writer skips the problem and jumps to features. PAS (Problem-Agitate-Solution) plus Proof and Action is a 50-year-old framework for a reason — it mirrors how readers decide. The "no imagine openers" rule kills one of AI's most-overused tropes. After the model drafts, cut 30% of the words by hand. It will be sharper.
You are a senior CRO strategist. Generate a structured A/B test hypothesis from the inputs below, complete with control description, variant description, primary metric, expected lift, sample size estimate, and risk of false positive…
Write a 350–450 word press release in AP style for the announcement below. Lead with the news in the first sentence, include a strong dateline, two quotes (founder + customer), a boilerplate, and a media contact block…
Draft a cold partnership email to a co-marketing prospect. Lead with shared audience overlap, propose three concrete co-marketing formats (webinar, swap, bundle) with a clear value-for-them, and end with a low-friction call to action…
Turn raw weekly metrics into an executive-ready summary. Output: 3-bullet TL;DR, what's working, what's not, top driver of change, one decision the leadership team needs to make, and three follow-up experiments ranked by expected ROI…