Most marketing prompt libraries are 90% padding and 10% prompts you'd actually use twice. This one is the opposite. Twenty-five prompts, grouped by job, written assuming you'll edit the output. They work in ChatGPT, Claude, or whichever model your team has standardized on.
This guide gives you the prompts and a short note on when each one shines — plus the staging pattern that makes any prompt work harder.
What changed in 2026
Two shifts: long context made full-funnel briefs feasible, and ad platforms exposed APIs that let you close the loop on what's working without leaving the model.
- Long context lets you paste full brand guidelines, past performance, and the brief in one shot.
- Ad platform APIs mean you can pull live performance into the prompt.
- Search distribution shift to Perplexity and ChatGPT changed top-funnel strategy.
How to use these
- Stage your prompts — research, brief, draft, edit. Not one giant ask.
- Include the audience and the desired action every time.
- Paste a winning past asset as a voice anchor.
- Ask for 3 variants, then merge.
- Save the keepers to a team prompt library.
1. Research prompts — best for cutting prep time
- SERP map. "List the top 10 results for query X. For each, note angle, length, and weakest section."
- Persona drill. "Build a persona for [audience] from these 5 customer transcripts. Real quotes, not summary."
- Objection list. "What 5 objections will this audience have to this offer? Rank by frequency."
- Competitor teardown. "Read these 3 competitor landing pages. What promises do they make? What do they avoid?"
- Trend scan. "Use Perplexity. What's been written about X in the last 30 days?"
2. Copy prompts — best with a voice anchor
- Headline barrage. "Here's a winning past headline. Write 15 in the same voice for this new offer."
- Landing page V1. "Brief: [paste]. Voice anchor: [paste]. Write H1, subhead, and 3 benefit bullets."
- About page rewrite. "Cut the corporate fluff. Rewrite in plain English. Keep the facts."
- CTA tester. "Write 10 CTA buttons for this page. Vary urgency."
- Tagline drill. "10 taglines for this product. Under 6 words. No 'reimagined' or 'next-gen.'"
3. Ad prompts — best for variant generation
- Variant grid. "Here's a winning ad. Write 20 variants — 5 each of pain, gain, social proof, urgency."
- Hook spinner. "Write 10 first-3-second hooks for a video ad on this offer."
- Image prompt brief. "Describe 5 image directions for this ad's hero. No people, brand colors X."
- Audience swap. "Same ad, rewritten for [B2B SaaS founders] vs [SMB owners]."
- Compliance pass. "Flag any claim in this ad copy that needs evidence."
4. Email prompts — best for sequence speed
- Welcome sequence. "Brief: [paste]. Write 5 emails over 2 weeks. Plain text. No 'I hope this finds you well.'"
- Subject line stack. "20 subject lines for this email. Aim for under 40 chars. Half curiosity, half value."
- Re-engagement. "Write a 3-email re-engagement sequence for users dormant 90 days."
- Cold outreach. "Write a cold email to [persona]. 80 words. One ask. No fake compliments."
- Newsletter intro. "Open this newsletter with a hook that earns the next 30 seconds."
5. Analysis prompts — best for closing the loop
- Performance summary. "Here's last month's ad data. What worked, what didn't, what to test next?"
- A/B postmortem. "Test results: [paste]. Why did B win? What's the next test?"
- Funnel diagnosis. "Funnel: 1000 → 300 → 50 → 5. Where do I lose people? What to fix first?"
- Voice of customer. "Cluster these 100 reviews into 5 themes. Pull a representative quote for each."
- Competitor move. "Competitor X just shipped Y. What's likely the strategic intent? How should we respond?"
Comparison: marketing AI stacks in April 2026
| Stack |
Cost |
Key strength |
Best for |
| Claude + Perplexity |
$40/mo |
Craft + research |
Solo / small team |
| ChatGPT Team |
$30/seat |
Shared prompt library |
5-50 person teams |
| Jasper |
$49/mo |
Brand voice training |
Multi-writer orgs |
| HubSpot AI |
bundled |
Native CRM/email |
HubSpot-heavy orgs |
Common mistakes to avoid
Vague audience. "Marketers" is too broad. "Series-A SaaS heads of marketing" is usable.
No voice anchor. Without one, you get the LinkedIn-influencer default voice.
Skipping the edit. Every prompt output needs a human pass before it ships.
FAQ
Which model is best for marketing prompts?
Claude for craft, ChatGPT for variants and structure, Perplexity for live research.
Can I share these prompts with my team?
Yes — and you should. A shared prompt library is a quiet competitive moat.
Are AI-written ads against platform rules?
Not generally. But platforms ban the obvious AI tells in some categories. Always edit.
Where to go next
For related guides see How to use AI for marketing in 2026, Best AI SEO tools in 2026, and How to build a content calendar with AI.