Most marketers don't need another list of "100 ChatGPT prompts." They need a handful of prompts that survive contact with a real deadline — a landing page due Friday, a welcome sequence that has to go out Monday, and a client asking for "the same post, but for LinkedIn and Telegram."
This is a working collection. Every prompt below has been used in production — for SaaS launches, e-commerce stores, and B2B service pages. Each one includes the exact structure, a realistic usage example, and notes on where models tend to hallucinate so you can catch it before your client does.
One principle before we start: a good marketing prompt is not a wish. It's a brief. The model needs a role, a constraint, a format, and a definition of "done." If you skip those, you get generic filler that reads like every other AI-generated landing page on the internet.
1. Positioning: The One-Sentence Value Proposition
Before writing a single headline, lock the positioning. This prompt forces the model to work from customer language, not marketing jargon.
You are a positioning strategist trained on the methodology from
April Dunford's "Obviously Awesome".
Product: [what it does, in plain words]
Target customer: [role, company size, context]
Top 3 alternatives they use today: [competitors or workarounds]
What we do better: [one honest differentiator]
Task:
1. Write 5 value proposition variants, max 12 words each.
2. For each, list the specific customer pain it addresses.
3. Flag any variant that could apply to a competitor unchanged —
those are too generic.
4. Pick the strongest and explain why in 2 sentences.
Example input: Product = "AI agent that automates CRM follow-ups for small sales teams." Alternatives = "manual follow-ups, Zapier, hiring a VA." Differentiator = "writes and sends follow-ups without a workflow builder."
The step that matters is #3. Without it, models default to "save time and money" — a phrase that describes literally every B2B tool ever sold.
2. Landing Page Headline Generator (With Anti-Cliche Filter)
Headlines are where AI output is most obviously AI. This prompt bakes in a blacklist.
Generate 15 landing page headlines for [product].
Rules:
- Max 10 words.
- Lead with a concrete outcome or a specific number.
- BANNED words: unlock, unleash, revolutionize, seamless,
game-changer, empower, next-level, supercharge.
- At least 5 must include a number, timeframe, or named metric.
- At least 3 must be written as a question the customer
actually asks themselves.
Output as a table: Headline | Type (outcome/question/number) |
Why it might work.
Run this twice with different temperature settings if your tool allows it. You'll usually find one usable headline in the first batch and a better one in the second.
3. Landing Page: Full Block-by-Block Structure
This is the prompt that replaces a blank Figma file. It generates the copy in the order a page actually reads.
Write landing page copy for [product] targeting [audience].
Structure (use these exact H2 headings):
1. Hero: headline + subheadline + CTA button text
2. Problem: 3 pain points in the customer's own words
3. Solution: how the product removes each pain (1:1 with above)
4. How it works: 3 steps, max 15 words per step
5. Proof: 2 placeholder case-study slots with [METRIC] brackets
6. Objection handling: 4 FAQ pairs
7. Final CTA
Constraints:
- Reading level: grade 8.
- No sentence longer than 20 words.
- Every claim that needs evidence gets a [SOURCE NEEDED] tag.
The [SOURCE NEEDED] tag is the honest part. The model will invent testimonials and stats if you let it. Tagging forces you to either find a real number or cut the claim.
4. Email Sequence: The 5-Email Welcome Flow
Welcome sequences are the highest-ROI automation most small teams never finish. This prompt produces a full draft in one pass.
Write a 5-email welcome sequence for [product/service].
Subscriber context: signed up for [lead magnet], has NOT purchased.
For each email give:
- Subject line (max 45 characters) + one A/B alternative
- Preview text (max 90 characters)
- Body (max 150 words)
- Single CTA
- Send delay (e.g., Day 0, Day 2)
Sequence arc:
Email 1 — deliver the promised value, set expectations
Email 2 — the problem most people get wrong
Email 3 — mini case study with a real-sounding scenario
Email 4 — address the top objection
Email 5 — direct offer with a deadline
Tone: [your brand voice, e.g., "direct, no hype, slightly dry humor"].
Where it breaks: the model loves fake urgency ("only 3 spots left!"). Delete any scarcity claim you can't actually honor. Fake deadlines are the fastest way to burn a list.
5. Subject Line Tester
Instead of asking for "good subject lines," ask the model to grade them.
Here are 10 subject lines for [email about X]:
[paste list]
Score each 1-10 on:
- Clarity (does it say what's inside?)
- Curiosity (does it create a reason to open?)
- Spam risk (words that trigger filters)
- Mobile truncation (fits in ~35 visible characters?)
Return a table sorted by total score, and rewrite the bottom 3.
This is faster than generating from scratch because you're using the model as a critic, which is where LLMs are genuinely strong.
6. Multi-Channel Repurposing: One Post, Ten Formats
This is the prompt that saves the most hours. It takes one core piece and rewrites it for every channel you actually use.
Take this source content: [paste article or post]
Rewrite it for these channels, one section each:
1. LinkedIn post (150-200 words, hook in first line, no hashtags)
2. X/Twitter thread (5-7 posts, each under 280 chars)
3. Telegram channel post (short paragraphs, one emoji max)
4. Instagram caption (under 125 words, CTA to comment)
5. Facebook post (conversational, question at the end)
6. Email newsletter blurb (80 words)
7. YouTube description (first 150 chars matter most)
8. Reddit comment (no self-promo, value-first)
9. Blog meta description (max 155 chars)
10. Push notification (max 40 chars)
Keep the core claim identical across all. Vary only tone,
length, and CTA.
Real usage: a 1,200-word blog post becomes a full week of channel content in about 4 minutes of generation plus 15 minutes of editing. The editing is non-negotiable — channel-native voice still needs a human pass.
7. Ad Copy Variants for Paid Campaigns
Paid ads need volume and testing discipline. This prompt generates a structured test matrix.
Write Google Ads and Meta ad copy for [product].
Target: [audience]. Primary pain: [pain]. Offer: [offer].
Google Ads (RSA):
- 15 headlines, max 30 characters each
- 4 descriptions, max 90 characters each
- Include the keyword [keyword] in at least 5 headlines
Meta:
- 5 primary texts (under 125 words)
- 5 headlines (under 40 characters)
- 3 CTAs from Meta's allowed list only
Group by angle: pain-led, benefit-led, social-proof-led,
objection-led, curiosity-led.
Character counts matter: Google Ads enforces 30/90 limits, and the model will happily exceed them unless you specify. Always verify lengths before upload.
8. SEO Brief Generator
A content brief that a writer or the model itself can execute against.
Create an SEO content brief for the keyword "[keyword]".
Include:
- Search intent (informational / commercial / transactional)
- Title tag (max 60 chars) and meta description (max 155 chars)
- Suggested H2/H3 outline covering the full topic
- 10 semantically related terms and questions to include
- People-Also-Ask style questions (5)
- Internal link opportunities (describe the topic, not a URL)
- Word count range based on competitor depth
Note: the model doesn't have live search volume data. Treat the brief as a structure, then validate keywords against a real tool like Google Keyword Planner or Ahrefs before committing.
9. Product Description for E-commerce
Product pages convert on specifics, not adjectives.
Write a product description for [product name].
Specs: [paste real specs]
Target buyer: [who and why they buy]
Top 3 competing products: [names]
Output:
- 1 short description (max 50 words) for category pages
- 1 long description (200-250 words) with a benefits block
- 5 bullet points, each starting with a benefit, not a feature
- 1 comparison line vs. each competitor (factual only)
Do NOT invent specs or certifications. If a claim isn't in the
input, omit it.
That last instruction prevents the classic failure mode: a model inventing "waterproof to IP68" for a product that is not.
10. Content Calendar for a Month
Build a 4-week content calendar for [brand] in [industry].
Mix per week:
- 2 educational posts
- 1 case study or proof post
- 1 behind-the-scenes
- 1 offer or CTA post
For each item give: day, channel, format (text/carousel/short
video idea), topic, hook line, and the CTA.
Align topics to these pillars: [pillar 1], [pillar 2], [pillar 3].
Pillars are the constraint that keeps the calendar from drifting into random tips. Without them, you get 20 disconnected posts.
11. Landing Page / Email Consistency Checker
After you've generated the landing page and the emails, run this to catch contradictions.
Here is my landing page copy:
[paste]
Here is my email sequence:
[paste]
Check for:
1. The main offer — is it described identically in both?
2. The price or terms — any mismatch?
3. The promised outcome — consistent?
4. Tone drift between assets
5. Any claim in one asset not supported elsewhere
Return a list of inconsistencies with the exact lines.
This catches the embarrassing case where your emails promise a free trial and your landing page says "book a demo."
12. A/B Test Hypothesis Generator
For this landing page [paste or describe], propose 6 A/B tests.
For each:
- Hypothesis in the format: "If we change X to Y, then Z
will improve because [reason]."
- Primary metric
- Estimated effort (low/medium/high)
- Priority score
Rank by impact-to-effort ratio.
Running this once gives you a testing backlog for a quarter. The hypothesis format matters — "change the button color" is not a hypothesis, it's a guess.
13. Brand Voice Extraction
If you have existing good copy, extract the voice so every future prompt stays consistent.
Analyze this writing sample: [paste 300-500 words of your
best-performing content]
Describe the brand voice as:
- 3 adjectives
- Sentence length pattern
- 3 words/phrases we use often
- 3 words/phrases we never use
- Formality level (1-10)
- How we address the reader (you/we/formal)
Then write a 4-line "voice card" I can paste into future prompts.
Paste that voice card at the top of every prompt in this list. It's the single highest-leverage habit for keeping AI output on-brand.
14. Objection Mining from Reviews
Real objections beat invented ones.
Here are 20 real customer reviews / support tickets: [paste]
Extract:
1. Top 5 objections or hesitations mentioned
2. Top 5 desired outcomes in customer's own words
3. 5 exact phrases worth reusing in ad copy
4. 3 things customers wish the product did
Quote the source text for each point.
This prompt turns your review section into a copywriting research database. The exact-phrase extraction is the most valuable part — those are the words your landing page should use.
A Note on What AI Still Can't Do
The prompts above handle drafting, structure, and repurposing. They do not handle:
- Real numbers. Every stat, testimonial, and metric must come from you. Models will fabricate statistics with total confidence — always verify before publishing.
- Legal claims. "Clinically proven," "guaranteed results," and regulated-industry claims need a human, ideally a lawyer.
- Actual customer truth. Positioning is only as good as the input. If your inputs are guesses, your output is polished guesses.
The workflow that works: use prompts to get to 70% in minutes, then spend your real time on the 30% that only a human who knows the customer can do.
Start With Three
You don't need all fourteen today. Pick the three that match your most urgent deadline: positioning plus headlines plus full page structure gets you a landing page. The welcome sequence plus subject line tester gets you an email flow. The repurposing prompt turns either one into a week of channel content.
Run them, edit hard, and keep the voice card handy — that's the difference between content that sounds like your brand and content that sounds like everyone else's AI.
Comments