50 ChatGPT Prompt Examples for Marketers (With Real Inputs, Outputs, and Why They Work)

Most "ChatGPT prompts for marketing" lists give you one-liners with no input, no output, and no idea why they work. You paste them, you get generic outputs, you assume the AI is the problem.

The AI is rarely the problem. The prompt is.

This guide includes 50 prompts organized by marketing task — content, SEO, email, social, ads, research, brand, and advanced frameworks. Every prompt comes with a real example input, a representative output, the rationale for why the prompt structure works, and a "watch out for" note when relevant.

Skim the sections below, jump to your task, and steal what you need.

Each prompt uses [BRACKETS] for the parts you swap. The brackets are placeholders — replace them with your actual context before sending. The more specific your replacements, the more useful the output.

Three principles run through every good prompt in this list. They are worth ten minutes before you start copying:

• Specificity beats length. A two-sentence prompt with three concrete constraints beats a six-paragraph prompt with vague instructions. "Write a 150-word LinkedIn post for B2B SaaS founders, opening with a counterintuitive statement" outperforms three paragraphs of context every time.
• Constraints beat instructions. Telling ChatGPT what NOT to do is more powerful than telling it what to do. "No jargon, no emoji, no rhetorical questions" produces more disciplined output than "make it professional."
• Examples beat descriptions. If you can show the AI a piece of writing in the style you want, do that. Two paragraphs of style sample teaches ChatGPT more than ten adjectives.

For deeper coverage, see How to Structure AI Prompts and Common Prompt Writing Mistakes. If you want to turn rough goals into structured prompts automatically, the Prompt Generator tool does that without signup.

Prompt 1: Blog post outline from a target keyword

Create a blog post outline for the target keyword "[KEYWORD]".
Audience: [AUDIENCE].
Goal: rank in the top 5 for this keyword.
Structure: H1, intro (2 sentences max), 5-7 H2s, FAQ section with 5 questions, conclusion.
For each H2, include a one-sentence description of what it should cover.
Make the H2s specific — no "Best practices" or "Tips and tricks" headings.

Example input: [KEYWORD] = "email subject lines for SaaS", [AUDIENCE] = "B2B SaaS marketers"

What good output looks like: ChatGPT returns an H1 like "Email Subject Lines for SaaS: 23 Examples That Get Opened in 2026" followed by H2s like "Why SaaS subject lines fail (and the pattern most teams repeat)", "Subject line formulas that work for cold outreach", "Subject line formulas for product launches", etc. — each with a sentence describing what the section covers.

Why it works: Three constraints stacked in one prompt — target keyword, audience, and a structural specification with the explicit "no generic headings" guardrail. The "rank in top 5" goal forces ChatGPT to think about competitor-style structure rather than freeform creativity.

Watch out for: ChatGPT defaults to listicle structures for almost everything. If you want a deeper analytical piece, add "this should be a thinking-piece, not a listicle" to the prompt.

Prompt 2: Listicle generation with real depth

Write a listicle blog post titled "[TITLE]" with exactly [N] items.
For each item, include:
- A specific, named example (real brand, real campaign, real number)
- One paragraph (60-90 words) of analysis
- One actionable takeaway
Avoid hedging language like "could", "might", "may be useful for".
Avoid filler items added to hit the count.

Example input: [TITLE] = "Cold Email Subject Lines That Got Above 60% Open Rates", [N] = 10

What good output looks like: Each item names a specific subject line, the company that sent it, and the actual open rate context. The analysis paragraph explains what made it work without generic statements.

Why it works: Forcing named examples kills the AI's natural tendency toward vague generalizations. The "no filler" instruction explicitly addresses the most common listicle failure — padding to hit a round number.

Watch out for: ChatGPT will sometimes invent fake examples to satisfy the "named example" rule. Always verify the specifics before publishing.

Prompt 3: Pillar page structure for a topic cluster

Design a pillar page structure for the topic "[TOPIC]".
Output:
1. The pillar page title and 6 H2 sections
2. For each H2, list 3-5 supporting blog posts that should exist to support this section (each as a working blog title)
3. Internal linking strategy: which supporting posts should link to which other supporting posts
4. Primary keyword target for the pillar page, plus 3 long-tail keywords per supporting post

Example input: [TOPIC] = "AI content detection"

Why it works: Explicitly asks for the hub-and-spoke architecture, including the linking strategy. Most marketers ask for "content clusters" and get back a list of titles with no relational logic. This prompt forces ChatGPT to plan the connections, not just the topics.

Related: See Keyword Clustering Explained for grouping the keywords this prompt produces.

Prompt 4: Content brief from a competitor URL

I want to write a post that competes with this article: [URL]
Analyze the article and produce a content brief that:
1. Identifies the target keyword and search intent
2. Lists every topic the competitor covers (as a flat list, no hierarchy)
3. Identifies 3-5 topics the competitor missed or covered poorly
4. Proposes a structure for my article that covers everything the competitor covers PLUS the gaps
5. Suggests an angle that differentiates my article without just being contrarian for the sake of it

Why it works: Many marketers ask ChatGPT to "outrank" a competitor and get back vague advice. This prompt converts the goal into a concrete editorial brief by separating coverage from differentiation.

Watch out for: If ChatGPT can't actually read the URL (older versions, no browsing enabled), paste the article text directly into the prompt instead.

Prompt 5: Repurpose blog post into newsletter

Convert this blog post into a 350-word email newsletter:
[PASTE BLOG POST]

Rules:
- Open with a single sentence that creates curiosity
- Pick the 2 most useful points from the post — leave the rest
- End with one specific takeaway the reader can act on this week
- No "Hello marketers" or "I hope this email finds you well" openers
- No links until the end

Why it works: Newsletters fail when they try to replicate the full blog post. The "2 most useful points" constraint forces ChatGPT to make editorial choices. The "no clichéd openers" rule cuts the AI tells.

Prompt 6: Repurpose blog post into LinkedIn carousel

Convert this blog post into a 7-slide LinkedIn carousel:
[PASTE BLOG POST]

For each slide:
- Slide 1: a hook that's either a counterintuitive statement, a specific number, or a one-sentence story
- Slides 2-6: one idea per slide, expressed as a 1-line headline + 2-3 supporting sentences
- Slide 7: a single takeaway and a CTA to read the full post

Output as a numbered list (Slide 1: ..., Slide 2: ...).
Keep each slide under 60 words.

Why it works: Most "repurpose to carousel" prompts produce 10-slide carousels with too much text per slide. Specifying 7 slides and a 60-word cap matches how LinkedIn carousels actually perform.

Prompt 7: Quarterly content calendar

Create a 12-week content calendar for [BRAND] in the [INDUSTRY] space.
For each week, include:
- Primary blog post topic (with target keyword)
- Two LinkedIn posts (with hooks)
- One newsletter theme
- One ungated lead magnet idea per month

Themes should follow a logical narrative — month 1 builds awareness, month 2 deepens trust, month 3 drives consideration.
No duplicate topics. No generic "industry trends" filler.

Example input: [BRAND] = "a CRM for solo founders", [INDUSTRY] = "small business SaaS"

Why it works: The three-month narrative arc instruction is what most calendar prompts miss. Without it, you get 12 random topics with no through-line.

Prompt 8: Topic ideation from a seed keyword

Generate 30 blog post ideas around the seed keyword "[KEYWORD]".
Split them into three categories:
- 10 ideas for top-of-funnel (educational, broad audience)
- 10 ideas for middle-of-funnel (comparison, decision-stage)
- 10 ideas for bottom-of-funnel (high-intent, ready-to-buy)

For each idea, include:
- The blog post title
- The likely search intent (informational, commercial, transactional)
- One sentence on why it would convert traffic into action

Skip ideas that are essentially the same topic with different wording.

Why it works: Funnel stage segmentation is the missing layer in most topic ideation. Asking for "30 ideas" gets you 30 similar ideas; asking for "10 each across three funnel stages" forces variety.

Prompt 9: Search intent analysis for a keyword

Analyze the search intent behind the keyword "[KEYWORD]".

Output:
1. Primary intent (informational / commercial / transactional / navigational) with one-sentence justification
2. The 3-5 sub-intents present in this query (what different things might users mean?)
3. The dominant SERP format (listicle, guide, tool, comparison, product page)
4. What the user actually wants to accomplish — not just what they typed
5. Two questions the user almost certainly has but didn't type

Why it works: "Search intent" is one of the most overused phrases in SEO content. This prompt forces specificity by asking for sub-intents and unspoken user questions — the parts that actually move rankings.

Related: The Search Intent Analyzer tool automates this for batch keyword lists.

Prompt 10: Title tag variations

Write 10 title tag variations for a blog post about "[TOPIC]".
Constraints:
- Under 60 characters each (count them)
- Include the primary keyword "[KEYWORD]" in 8 of the 10 (vary placement)
- Mix formats: 3 listicles (numbered), 3 question-format, 2 how-to, 2 with year/2026
- No clickbait. No "Ultimate Guide" or "Everything You Need to Know"
- Each variation should target a slightly different sub-intent

Why it works: Specifying character counts, formats, and forbidden phrases produces title tags you can actually A/B test, not just pick the first one.

Related: The Meta Title Generator automates this with character counts.

Prompt 11: Meta description generation

Write 5 meta descriptions for this URL: [URL] (or describe the page if no URL available)

Constraints:
- 150-160 characters each
- Include the primary keyword "[KEYWORD]" naturally
- End with an implicit or explicit benefit (what the reader gets)
- No "Click here" or "Read more"
- Each one should test a different psychological angle: curiosity, urgency, specificity, social proof, problem-statement

Why it works: The five-different-angles instruction prevents the AI from producing five near-identical descriptions, which is what most prompts return.

Prompt 12: Featured snippet targeting

I want to rank in the featured snippet for the query "[QUERY]".

Look at the typical snippet format for queries like this (paragraph, list, or table?) and produce:
1. A 40-60 word definition-style answer formatted exactly as Google's paragraph snippets render
2. A list-style version (5-7 items, each 1 line)
3. A table version if applicable (2-3 columns)

For each, explain which version is most likely to be picked based on current SERP behavior for this query type.

Why it works: Most snippet-targeting prompts just say "write a featured snippet." Asking for all three formats and the rationale gives you something you can A/B in different sections of your post.

Prompt 13: Internal linking suggestions

I have a blog post about "[NEW POST TOPIC]".
Here's my existing site's content inventory:
[PASTE list of 10-20 existing post titles + URLs]

Suggest:
1. 5 existing posts I should link FROM in my new post (with the suggested anchor text)
2. 3 existing posts I should add a link TO my new post FROM (with anchor text)
3. Any clusters where my new post is the missing piece

Why it works: Internal linking prompts usually produce vague advice. Pasting your real content inventory turns the prompt into a structured recommendation.

Related: The Internal Linking Tool automates this at scale.

Prompt 14: Schema markup generation

Generate JSON-LD schema markup for this content type: [TYPE — e.g., HowTo, FAQPage, Article, Product, Review]

Content details: [PASTE the relevant content]

Output:
1. The complete JSON-LD code block, properly formatted
2. A brief note on which fields are required vs optional in this schema type
3. Common implementation mistakes for this schema type

Watch out for: Always validate the output in Google's Rich Results Test before deploying. ChatGPT occasionally invents fields that don't exist in schema.org.

Prompt 15: Keyword clustering

Group these keywords into clusters based on shared search intent:
[PASTE keyword list — typically 20-100 keywords]

For each cluster:
- A cluster name
- The primary keyword (highest-volume or clearest representation of the intent)
- The other keywords in the cluster
- A one-sentence description of the user intent
- The recommended page type to target the cluster (blog post, tool, landing page, comparison)

Flag any keywords that don't fit well into any cluster.

Related: The Keyword Cluster Generator does this with intent-grouping and full keyword lists. See also Keyword Clustering Explained.

Prompt 16: Cold outreach email

Write a cold outreach email to [TARGET ROLE] at [TARGET COMPANY TYPE].
Purpose: [WHAT YOU WANT FROM THEM]

Constraints:
- Under 100 words total
- Personalized opener (use this hook: [SOMETHING SPECIFIC ABOUT THE PROSPECT OR THEIR COMPANY])
- One specific value proposition tied to their likely current pain
- Single clear CTA (specific time/length, not "let me know if interested")
- No "I hope this email finds you well", no "Just following up", no "Quick question"

Example input: [TARGET ROLE] = "Head of Content", [TARGET COMPANY TYPE] = "B2B SaaS Series A", [PURPOSE] = "explore a partnership where we co-create a comparison article"

Why it works: Cold email AI is notorious for producing template-sounding emails. The forbidden-phrase list strips out the most common AI tells. The single-CTA rule prevents the "or also we could..." trailing offers that kill response rates.

Related: The Cold Email Writer tool and Cold Email Mistakes That Kill Reply Rates.

Prompt 17: Subject line A/B test variations

Write 10 email subject line variations for an email about [TOPIC].
The email's main value: [ONE-LINE VALUE PROP]
Audience: [AUDIENCE]

Constraints:
- Under 50 characters each
- 3 should be question-based
- 3 should be specific/number-based ("3 reasons", "$2,400")
- 2 should be curiosity-gap
- 2 should be direct value-statement
- None should use the word "amazing", "ultimate", "secrets", or emoji

Related: The Email Subject Generator and Cold Email Subject Lines for Sales.

Prompt 18: Welcome email sequence

Write a 5-email welcome sequence for new [PRODUCT TYPE] signups.
Audience: [AUDIENCE]

For each email, include:
- Subject line
- Send timing (day 0, day 1, etc.)
- Goal of this specific email
- 100-150 word body
- Specific CTA

Sequence arc:
- Email 1 (day 0): warm welcome + immediate quick win
- Email 2 (day 1): one specific feature with a real example
- Email 3 (day 3): social proof (case study or specific result)
- Email 4 (day 7): handle the most common objection or concern
- Email 5 (day 14): invite to a paid plan / next step with a clear deadline

Why it works: Without specifying the narrative arc per email, sequence prompts produce five emails that all sound the same. The per-email goal forces variety.

Prompt 19: Re-engagement email

Write a re-engagement email for [SEGMENT] who haven't opened our emails in [TIMEFRAME].

Constraints:
- Subject line under 40 chars
- Body under 80 words
- Acknowledge the silence without guilt-tripping
- Offer one specific reason to re-engage (not "we miss you")
- Single CTA: either "stay subscribed" link or "unsubscribe" link (don't bury either)

Why it works: Most re-engagement prompts produce passive-aggressive "we miss you" emails. The "no guilt-tripping" constraint forces a different angle.

Prompt 20: Newsletter teaser

Write a 60-word teaser for our weekly newsletter.
This week's main story: [TOPIC]
Key insight: [INSIGHT]

The teaser will be used:
- As the preview text in the email itself
- As a LinkedIn post pointing to the signup page
- As a tweet

Output 3 versions optimized for each context. Each should hint at the insight without giving it away.

Why it works: Asking for three context-specific versions in one go produces variants you can actually use across channels.

Prompt 21: Sales follow-up email

Write a follow-up email after a sales call.
Context: [WHAT WAS DISCUSSED]
Specific next steps agreed: [ACTIONS]
Prospect's main concern: [CONCERN]

Constraints:
- Under 120 words
- Reference one specific thing from the conversation (use this detail: [SPECIFIC DETAIL])
- Address the concern directly but briefly
- Confirm the next step with a specific date/time
- No "It was great connecting!", no "Just circling back"

Prompt 22: Customer success email

Write an email to a customer who has just hit [MILESTONE] with our product.

Goal: celebrate the milestone, surface one underused feature that would help them go further, and ask for feedback or a testimonial — but don't make it feel like a marketing email.

Constraints:
- Under 100 words
- Genuine tone, no jargon
- Single primary CTA (testimonial OR feature exploration, not both)
- Personalization: include [THEIR USE CASE / SPECIFIC RESULT]

Prompt 23: LinkedIn thought leadership post

Write a LinkedIn post about [TOPIC] from the perspective of [ROLE/EXPERTISE].

Structure:
- Hook: 1 sentence that's either a counterintuitive claim, a specific number, or a one-line story
- Setup: 2-3 sentences of context (max 50 words)
- The actual insight: 3-5 short paragraphs (1-3 sentences each) with line breaks between each
- A specific example or anecdote
- A closing question that invites engagement (not "thoughts?" — something specific)

Tone: confident but not arrogant. Personal but not narcissistic.
Avoid: emoji, "I'm excited to share", "Here's what I learned", motivational closing lines.

Why it works: LinkedIn's algorithm rewards posts with high dwell time and meaningful comments. This structure is designed for both: a strong hook to stop the scroll, a real insight in the middle, and a specific closing question (not generic "what do you think?").

Prompt 24: Twitter/X thread

Write an X thread about [TOPIC]. Total length: [N] tweets.

Tweet 1: the hook. Must work as a standalone tweet. State the conclusion or the most provocative point upfront.

Tweets 2 to [N-1]: one idea per tweet. 200-260 characters each (not 280 — leave breathing room).

Final tweet: either a summary, a contrarian flip, or a CTA.

Rules:
- Each tweet should be readable in isolation
- No "1/" "2/" numbering (the thread structure shows this)
- No emoji unless absolutely critical to the meaning
- One specific number or named example minimum across the thread

Related: How to Write Viral Tweets covers the structural patterns underneath this prompt.

Prompt 25: Instagram caption with hook

Write an Instagram caption for a post about [TOPIC].

Structure:
- First line: the hook (this is all that shows before "...more"). Must be intriguing on its own.
- Second line: blank
- 3-5 short paragraphs, 1-2 sentences each, with line breaks between
- Closing: a single question or CTA
- Hashtags: 8-12 relevant ones, mixed between high-volume and niche

Tone: [TONE — e.g., conversational, expert, playful]
Audience: [AUDIENCE]

Related: The Hook Generator tool for the first-line problem specifically.

Prompt 26: TikTok / short-form video script

Write a 45-second TikTok script about [TOPIC].

Structure:
- Hook (first 3 seconds): a question, claim, or visual cue that stops the scroll
- Setup (seconds 3-10): why this matters
- Payoff (seconds 10-35): the actual content / 3 points / story
- Outro (seconds 35-45): the takeaway + a soft CTA (follow for more, comment your X, etc.)

Output format:
[Timestamp] [Visual/action description] | [Spoken/text on screen]

Constraints:
- Conversational, not script-y. Read it out loud — if it sounds written, rewrite.
- No "Today I'm going to talk about..."

Prompt 27: Repurpose long-form into 5 social posts

Convert this long-form content into 5 different social media posts:
[PASTE THE LONG-FORM CONTENT]

Output:
- 1 LinkedIn post (300 words, professional angle)
- 1 X thread (5 tweets, contrarian or insightful angle)
- 1 Instagram caption (with hook + hashtags)
- 1 short-form video script (45 sec)
- 1 newsletter teaser (60 words)

Each one should pull a DIFFERENT angle or insight from the source. Don't just rewrite the same point five times.

Why it works: "Don't just rewrite the same point five times" is the constraint that turns generic repurposing into actually-different content.

Prompt 28: Social media bio

Write 5 different versions of a [PLATFORM] bio for someone who is [ROLE/DESCRIPTION].
Their main offer/value: [VALUE]
Audience they want to attract: [AUDIENCE]

Each version should test a different angle:
1. Credentials-led (positions through expertise)
2. Outcome-led (positions through results delivered)
3. Personality-led (positions through tone/perspective)
4. Audience-led (positions through who they help)
5. Contrarian (positions against a common belief in their space)

Each under [PLATFORM character limit]. Include line breaks where the platform allows.

Related: The Bio Generator handles platform-specific character limits automatically.

Prompt 29: Google Ads headlines and descriptions

Write Google Ads copy for a campaign promoting [PRODUCT/OFFER].
Target keyword: [KEYWORD]
Audience: [AUDIENCE]
Landing page promise: [WHAT THE LANDING PAGE DELIVERS]

Output:
- 15 headlines (each under 30 characters — count them)
- 4 descriptions (each under 90 characters)
- 2 of the headlines should include the exact target keyword
- Headlines should vary by angle: 3 benefit-led, 3 feature-led, 3 social-proof, 3 question-form, 3 urgency/scarcity

Each must be runnable in isolation — Google Ads rotates these.

Why it works: The character counts and angle distribution match Google's responsive search ads requirements. The "runnable in isolation" rule prevents the AI from writing headlines that only work together as a sequence.

Prompt 30: Facebook/Meta Ads primary text

Write Facebook Ads primary text for [PRODUCT/OFFER].
Target audience: [AUDIENCE]
Audience awareness level: [unaware / problem-aware / solution-aware / product-aware / most-aware]
Funnel stage: [top / middle / bottom]

Output 5 variations:
- 100-150 words each
- Each must hook in the first 125 characters (before the "See more" cut on mobile)
- Vary the opening: story, question, statistic, contrarian statement, direct callout

Avoid: emoji at the start, fake urgency ("limited time!"), generic "transform your business" language.

Why it works: Specifying awareness level is what separates this from generic ad copy prompts. Copy for "unaware" audiences must do education first; copy for "most-aware" audiences must do conversion first. The same product needs different ads at different stages.

Prompt 31: Display ad copy with CTA variations

Write display ad copy for [PRODUCT].
Ad size context: [responsive / specific sizes]
Brand voice: [DESCRIBE OR PASTE EXAMPLE]

Output:
- 5 headline options (under 25 characters each)
- 5 description options (under 90 characters each)
- 8 CTA button text options (under 15 characters each)

CTAs should range from low-commitment ("Learn More") to high-commitment ("Buy Now"). Label each CTA with its commitment level.

Prompt 32: Retargeting ad with social proof

Write retargeting ad copy for visitors who [SPECIFIC ACTION TAKEN — e.g., viewed pricing but didn't sign up].

Use this social proof: [REAL TESTIMONIAL / STAT / CASE STUDY]

Constraints:
- Acknowledge they've seen us before (without being weird about it)
- Lead with the social proof
- Address the most likely objection from their behavior (didn't convert because of price, complexity, or trust?)
- Single CTA matched to where they were in the funnel
- Multiple versions for each of the three likely objections

Prompt 33: YouTube ad script

Write a YouTube ad script for [PRODUCT/OFFER]. Length: [15 / 30 / 60 seconds].

Structure:
- Hook (first 5 seconds): the audience must want to keep watching. Pattern interrupt, specific claim, or named pain.
- Body: introduce the product/offer and the core benefit. One specific result or stat.
- CTA: clear, specific, with a friction-reducing element

Format the output as: [Timestamp range] [What happens on screen] | [Voiceover/spoken text]

Avoid: "Hi, I'm [name] and today I want to talk about...", testimonial montages, generic "transform your life" promises.

Prompt 34: Competitor positioning analysis

Analyze the positioning of these competitors:
1. [COMPETITOR 1] — their homepage hero copy: [PASTE]
2. [COMPETITOR 2] — their homepage hero copy: [PASTE]
3. [COMPETITOR 3] — their homepage hero copy: [PASTE]

For each:
- The audience they're positioning toward
- The specific job-to-be-done they're solving
- The implicit "vs alternative" — what are they positioning AGAINST
- The unique angle (or lack of one)

Then:
- Where are the gaps in this market's positioning collectively?
- What angle would differentiate from all three?

Why it works: Pasting actual hero copy (not just company names) gets you analysis grounded in real positioning, not ChatGPT's training-data assumptions.

Prompt 35: Customer persona generation

Build a detailed customer persona for [PRODUCT/SERVICE].

Real customer data I have (or best guesses):
- Typical job title: [TITLE]
- Industry: [INDUSTRY]
- Company size: [SIZE]
- Stated goals: [WHAT THEY SAY THEY WANT]
- Real pain we've observed: [WHAT WE'VE LEARNED FROM CALLS / SUPPORT]

Output:
- Name and one-sentence summary
- 3 specific anxieties (not generic — based on their real role)
- 3 specific goals (career/business outcomes, not "make more money")
- The 3 questions they ask before buying
- The 3 internal stakeholders they need to convince
- Their typical buying journey: where they start research, what they need to see, what kills the deal

Skip demographic filler (age, hobbies) unless directly relevant.

Why it works: Most persona prompts produce demographic filler. This one focuses on the buying-relevant detail and asks for the internal-stakeholder layer that B2B marketers actually need.

Prompt 36: Survey question generation

Design a customer survey to learn about [SPECIFIC LEARNING GOAL].
Target respondent: [WHO]
Channel: [HOW THEY'LL TAKE IT — email, in-app, exit intent, etc.]
Length constraint: [TIME — 2 min, 5 min, etc.]

Output:
- The exact questions in order
- The format for each (multiple choice, scale, open-ended) and why
- Skip-logic where relevant
- The single most important question (the one we can't lose if we have to cut)

Avoid leading questions, double-barreled questions, and survey fatigue patterns.

Prompt 37: Review/feedback synthesis

Synthesize this collection of customer reviews/feedback:
[PASTE 10-50 reviews]

Output:
- Top 5 themes (with frequency)
- Most common specific praise
- Most common specific complaint
- Quotes that best represent each theme (use actual quotes from the input, don't paraphrase)
- The "unsaid" pattern — what isn't mentioned but should be, based on the gaps
- One actionable insight per theme

Why it works: "Don't paraphrase" forces ChatGPT to find real quotes, which surfaces voice-of-customer language you can use in copy.

Prompt 38: Trend analysis for a niche

I want to identify emerging trends in [NICHE].

Here's what I'm seeing in my own research: [PASTE OBSERVATIONS — articles, conference talks, social posts you've noticed]

Help me:
1. Group these observations into 3-5 distinct trend categories
2. For each trend, assess: is this a real shift or a hype cycle?
3. Identify which trends create marketing opportunities for [YOUR BUSINESS] specifically
4. Flag any trends that COMPETE with our current positioning

Be skeptical. Push back on trends that look thin.

Why it works: "Be skeptical" + "push back on trends that look thin" is the constraint that prevents ChatGPT's natural tendency to validate everything you say.

Prompt 39: SWOT analysis

Conduct a SWOT analysis for [BRAND/PRODUCT].

Context:
- Current positioning: [POSITIONING]
- Main competitors: [COMPETITORS]
- Recent wins: [WINS]
- Recent struggles: [STRUGGLES]

Output:
- 4-6 items per quadrant
- Each item should be specific (not "strong brand" — say what makes it strong)
- For each Weakness, suggest one concrete intervention
- For each Threat, assess likelihood (high/medium/low) and time horizon (immediate/12mo/3yr)

Then: the top 3 strategic priorities that emerge from this analysis.

Prompt 40: Brand voice guidelines from existing copy

Reverse-engineer brand voice guidelines from this existing content:
[PASTE 3-5 representative pieces of brand content — homepage, blog excerpts, social posts]

Output:
- The brand voice in 3 adjectives (with examples from the input proving each)
- Sentence length tendencies
- Vocabulary patterns (words it favors, words it avoids)
- How it handles humor (or doesn't)
- How it handles weakness/vulnerability
- 5 "do" rules
- 5 "don't" rules
- A 100-word sample written in the voice for a topic the original content didn't cover, to verify the voice transfers

Why it works: The "100-word sample on a new topic" is the verification step. If the AI gets the voice right on a new topic, the guidelines are real. If not, they're hallucinated.

Prompt 41: Tagline generation

Generate 15 tagline options for [BRAND].
Brand essence: [WHAT THE BRAND STANDS FOR]
Target customer: [WHO]
Differentiator: [WHAT MAKES THIS DIFFERENT]

Variation:
- 5 outcome-focused (what the customer gets)
- 5 identity-focused (who the customer becomes)
- 5 contrarian (positions against the industry norm)

Constraints:
- Under 8 words each
- Memorable on first read
- No "Empowering [X]", no "Unleash your [Y]", no industry clichés
- Each must work without context (would survive on a billboard alone)

Prompt 42: Product naming

Generate 20 name options for [PRODUCT DESCRIPTION].

Variation:
- 5 descriptive names (says what it does)
- 5 evocative names (says how it feels)
- 5 invented words (memorable, ownable, easy to say)
- 5 references (mythology, metaphor, cultural touchpoint)

For each, include:
- The name
- A one-line rationale
- A potential downside (sounds like X, hard to spell, etc.)

Skip names that are already trademarked obvious brands.

Prompt 43: Brand story / about-us page

Write an about-us page for [BRAND].

Inputs:
- Founder origin: [WHY THIS WAS BUILT]
- What we believed at the start: [FOUNDING BELIEF]
- What we've learned since: [WHAT CHANGED]
- What we believe now: [CURRENT BELIEF]
- The customer we're really for: [SPECIFIC AUDIENCE]

Structure:
- Open with a tension or contradiction, not a feature list
- Tell the founding story in 2-3 paragraphs (specific, not generic)
- Articulate the belief / point of view (this is what differentiates)
- Connect to who the customer is and what we're trying to help them do
- End with a forward-looking statement, not a CTA

Avoid: "Founded in [year]", "Our mission is to...", lists of features, generic vision statements.

Prompt 44: Customer testimonial framework

Write a customer testimonial request email and a follow-up framework.

Email: asking [SPECIFIC CUSTOMER] for a testimonial after [SPECIFIC SUCCESS THEY HAD].

Then: a framework for the actual testimonial:
- 3 questions that produce specific, story-based answers (not generic "I love this product")
- 1 question that surfaces the doubt they had before buying (this becomes the most useful testimonial content)
- 1 question that surfaces the specific outcome (with a number if possible)

Output the email AND the testimonial questions. The email should mention they'll only have to answer 5 short questions.

The CRISPE framework — Capacity & Role, Insight, Statement, Personality, Experiment — is one of the most reliable structures for producing high-quality strategic output from ChatGPT. See AI Prompt Frameworks for the full method. The five prompts below show CRISPE applied to specific marketing tasks.

Prompt 45: Comprehensive content strategy (CRISPE)

Capacity & Role: Act as a content strategist with 10 years of experience helping B2B SaaS companies in the [INDUSTRY] space build organic growth engines from scratch.

Insight: I'm the [ROLE] at [COMPANY], a [STAGE — early/growth/scale] company that sells [PRODUCT] to [AUDIENCE]. Current traffic: [NUMBER]. Current conversion rate: [NUMBER]. Our main constraint is [CONSTRAINT].

Statement: Build a 6-month content strategy that prioritizes content with the highest probability of moving the conversion rate, given our current traffic constraints. Include topic clusters, pillar pages, and a sequencing rationale.

Personality: Direct, strategic, willing to tell me where my current approach is wrong. Skip the encouragement. Push back on assumptions you think are flawed.

Experiment: First produce the strategy. Then, in a separate section titled "Alternative angles I considered and rejected," explain two other strategic directions you considered and why this one wins.

Prompt 46: Multi-channel campaign brief (CRISPE)

Capacity & Role: Act as a senior integrated campaign strategist who has run launches for products at every stage from pre-launch to scale.

Insight: We're launching [PRODUCT] on [DATE]. Our budget is [BUDGET]. Target audience: [AUDIENCE]. Primary KPI: [KPI]. We have [CHANNELS AVAILABLE]. Our biggest risk is [RISK].

Statement: Create a launch campaign brief covering the 4 weeks before launch, launch week, and 4 weeks after. For each phase, specify channel mix, key messages, content needed, and success metrics.

Personality: Practical, opinionated, prioritizes high-leverage actions over comprehensive coverage.

Experiment: Before writing the brief, ask me 3 clarifying questions that would meaningfully change the recommendation. Don't ask anything I've already answered.

Prompt 47: Customer journey mapping (CRISPE)

Capacity & Role: Act as a customer experience strategist specializing in [INDUSTRY], with deep expertise in mapping the actual (not idealized) journey customers take.

Insight: My customers are [DESCRIPTION]. They typically discover us via [CHANNELS]. The friction points we've identified so far are [FRICTION]. We have data on [WHAT DATA YOU HAVE].

Statement: Map the customer journey from first awareness to power user, identifying the 5 most critical friction points and the marketing/product/content interventions that would unblock each.

Personality: Empathetic to the customer but skeptical of internal assumptions. Distinguish between what we WISH happens and what ACTUALLY happens.

Experiment: After the map, identify the single intervention that, if successful, would unlock the most value. Explain your reasoning, including what would need to be true for you to be wrong.

Prompt 48: Brand audit (CRISPE)

Capacity & Role: Act as a brand strategist who has audited brands across multiple industries and stages, known for honest assessments rather than diplomatic ones.

Insight: My brand is [BRAND]. Here's a representative sample of our current expression: [PASTE: homepage hero, About page intro, recent social posts, a blog post intro]. Our intended positioning is [POSITIONING].

Statement: Audit the gap between our intended positioning and our actual brand expression. Identify the 5 most significant inconsistencies and 3 patterns that are working well.

Personality: Honest to the point of bluntness. The kind of feedback a friend with strong taste gives, not the kind a paid consultant gives.

Experiment: At the end, propose two possible repositioning angles: one that minimally adjusts current brand to fix the gaps, and one that more aggressively claims a stronger position. Compare the risk/reward of each.

Prompt 49: Marketing experiment design (CRISPE)

Capacity & Role: Act as a growth experimentation lead who has run hundreds of marketing tests and knows the difference between an insight-generating experiment and a waste of time.

Insight: I want to test [HYPOTHESIS]. Current state: [BASELINE METRICS]. Our traffic volume: [TRAFFIC]. The cost of running the test: [COST].

Statement: Design the experiment. Specify the test variant, control, success metric, secondary metrics to monitor, sample size needed, expected runtime, and the decision criteria for what we'll do based on each possible outcome.

Personality: Rigorous, skeptical, willing to tell me the test is a bad idea if it is.

Experiment: Before designing the test, evaluate whether the hypothesis is worth testing. If it's likely to produce no actionable insight regardless of outcome, say so and recommend a better hypothesis.

Prompt 50: Improve any prompt

Here's a prompt I've been using: [PASTE YOUR PROMPT]

Critique it. Specifically:
1. What's vague where it should be specific?
2. What's instructed where it should be constrained?
3. What's missing? (audience, format, length, examples, things to avoid)
4. What's redundant or contradictory?
5. Rewrite it with the fixes.

Then: run both the original and your rewrite against this test case: [TEST CASE]. Show both outputs side by side.

This is the most useful prompt on this page. Run any of the previous 49 through it before you commit to using them in production.

The patterns that don't work

After the 49 prompts above, here are the marketing prompts that consistently underperform — and what to do instead.

• "Be creative" — ChatGPT has no model of what "creative" means to you. Instead: "Write three openings: one shocking, one funny, one data-driven."
• "Write a viral post" — AI can't predict virality. Instead: "Write a post structured like this one that went viral: [paste actual example]."
• "Make it engaging" — Undefined. Instead: "Open with a question, include one specific number, end with a question for comments."
• "Write like [famous person]" — AI does shallow imitation by mimicking surface tics. Instead: "Here are 2-3 paragraphs of their actual writing: [paste]. Match this voice."
• "Make this better" — Better how? Instead: "Make this shorter (target 60% of current length), more specific (add at least two concrete numbers), and less hedging (remove 'might' and 'could')."
• "Generate ideas" — Will get you generic ideas. Instead: "Generate ideas. For each one, also generate one reason it might not work and one specific test we could run cheaply to validate it."

Notice the pattern in every prompt that works: they shift ChatGPT from being asked to do creative output to being asked to follow structured constraints.

Bad prompts ask the AI to write. Good prompts ask the AI to fill in a template you've already designed.

The marketers getting outsized value from AI aren't using better tools. They're using clearer instructions. Every prompt above is essentially a constrained template with three to five guardrails. Once you can write your own version of these from scratch, you don't need to copy prompts anymore — you can build them for any task.

If you want help getting to that point, the Prompt Generator tool turns rough goals into structured prompts, and How to Structure AI Prompts covers the underlying patterns in depth.