At a glance
Most marketing teams are using AI for output volume and not asking it the better question: what would actually move this customer? The eight prompts below cover campaign strategy, brand voice, audience research, landing-page copy, email sequences, social content, ad creative, and repurposing. Each prompt is short. Each one is paired with the constraint that prevents the most common AI failure for that specific marketing job. The output of these prompts is not finished work; it is the strongest starting point you can give a human marketer who has 90 minutes to ship something.
Why does AI-driven marketing copy land flat so often?
The marketing-AI failure mode that hurts most brands is not bad output. It is uniformly average output. AI without constraint produces copy that pattern-matches across every channel a customer touches before they convert. The LinkedIn post, the landing page, and the welcome email share the same rhythm, the same em-dash habit, the same triple-adjective phrases. That sameness is what makes “AI-driven marketing” feel like noise. The fix is not to abandon AI; it is to feed it specifics (brand voice samples, audience research, the exact objection to address) so the work it produces lands instead of dilutes. The marketer who supplies real inputs gets work that converts. The marketer who supplies “write me a landing page for X” gets the average of every landing page on the internet.
The eight prompts below all share the same shape: a job to do, the inputs that make AI capable of doing it, the constraints that block the failure mode, and the output format that makes the result usable by a human reviewer. Together they cover the marketing functions where AI is most useful (and most often misused).
⚠️ The 2026 marketing-AI hazards to know
LinkedIn now down-ranks posts flagged as AI-generated (rollout confirmed October 2025 when LinkedIn's VP of Global Editorial Laura Lorenzetti told Entrepreneur). Google's AI Overviews change conversion math: top-of-funnel content gets summarized into the result page, and click-through-rate on summarized queries drops 30-50%. Brand-voice drift compounds across channels; what looks fine on a single asset reads as scattershot across the customer journey. The prompts below have constraints aimed at each of these. Pair with our How to Edit AI Out of Your Writing guide before any public-facing copy ships.
What are the eight marketing prompts?
1. Campaign strategy
Here is the business goal: [paste, with a specific metric and a deadline].
Help us design the campaign:
– The single customer belief that has to change for them to convert
– The 3 channels where this customer is reachable (with evidence from the transcript, not a guess)
– The hook for each channel that speaks to that belief
– The next-step CTA that matches the channel and the belief
For each recommendation, cite the specific sentence from the customer transcript that supports it.
Why this works: The “single customer belief that has to change” framing is the question that most marketing teams skip and that determines the campaign’s effective conversion ceiling. Asking AI to ground every recommendation in a transcript quote prevents the hallucinated-persona problem where the AI invents a customer who buys easily. The transcript is the difference between marketing strategy and marketing wishful thinking.
2. Brand voice extraction
Analyze the voice across these samples. Output:
– 5 voice characteristics with one specific example from the samples each
– 5 sentence-rhythm patterns we use (length, punctuation, sequence)
– 10 words or phrases we use that are unusual or distinctive
– 5 words or patterns we DO NOT use that competitors do
Then build a 1-page voice brief any team member could hand to AI for future copy work. Format: rules with examples.
Why this works: “Match our brand voice” is the request AI fails at most often because most teams have never written down what their brand voice actually is. This prompt does the work of extracting it from existing copy, then turns it into a portable artifact. The voice brief becomes a system prompt for every later prompt; it is the single highest-impact investment a marketing team can make in their AI workflow.
3. Audience research synthesis
Synthesize:
– The 5 most common things customers say in their own words about the problem we solve
– The 3 most common objections that block purchase
– The 3 most common reasons they leave for a competitor
– The phrases customers use that we should mirror in our copy
– The phrases we use that nobody actually uses about us (the ones to retire)
Quote the source on every claim. Do not invent quotes.
Why this works: Real customer language is the single most underused input in marketing. Most teams write copy that says what they wish their customers said; this prompt outputs what they actually say. The “phrases we use that nobody actually uses” output is uncomfortable and the most useful: it is where the AI tells you which copy lines to kill.
4. Landing page copy
– Voice brief: [paste from prompt 2]
– Audience pain in customer language: [paste from prompt 3]
– The single belief that has to change: [paste from prompt 1]
– The CTA we want: [name the action]
Structure: hero (8-12 word headline + one-sentence subhead), 3 problem statements in customer language, 3 reframes that map our offer to those problems, social proof block (leave [PROOF NEEDED] placeholders), pricing/CTA, FAQ with 5 real questions from the research.
Constraints: no "in today's fast-paced world", no "synergy", no rhetorical-question headlines. Match the voice brief. Use customer language verbatim where possible.
Why this works: This is the prompt that becomes possible only after the first three. Most teams skip ahead to “write me a landing page” and get the average internet landing page. With brand voice, customer language, and the belief change locked in, the AI output goes from generic to specifically yours. The [PROOF NEEDED] placeholder is the safety: AI will invent testimonials and stats if you let it; the placeholder forces a human to verify.
5. Email nurture sequence
Inputs: voice brief, audience research, the 3 most common objections.
For each email:
– One subject line under 7 words
– One sentence opening (no "hope this email finds you well")
– The angle (one objection or one belief to address; each email picks a different one)
– The structure (story, value, ask, social proof, or CTA)
– The CTA (only one per email)
– The send delay after the previous email
Hard rule: each email is 90-120 words. Each one earns the next one. Email 5 makes the strongest single ask of the sequence.
Why this works: Most AI email sequences sound like every other AI email sequence and the data shows it: the average reply / convert rate on automated AI sequences in 2025 hovered 20-30% below human-written. The 90-120 word constraint and the “each email picks a different objection” rule prevents the repetitive-pacing failure mode. One CTA per email enforces the discipline that AI defaults to violating.
6. Social content with voice fidelity
Each post should:
– Open with a specific claim or scene (not a question, not "here are 5 ways")
– Be one screenful on phone (LinkedIn: 3 short paragraphs max; X: under 280; Threads: similar to X)
– Avoid: em-dashes, triple-adjective phrases, "here's the thing", rhetorical question openers, "in this thread we will"
– Include one specific small detail that only a human in our company would know
For each, tell me which voice characteristic from the brief it expresses most. Mark which one is most likely to be flagged by LinkedIn's AI-content detector and what to change.
Why this works: The “one specific small detail only a human in our company would know” is the line that makes social posts unmistakably yours. LinkedIn’s 2025 AI-content detection is real and the consequence (down-ranking, soft suppression in feed) is silent; asking AI to self-flag which post is most exposed is the single best inoculation against the algorithmic penalty.
7. Ad creative variations
Variation structure:
– 5 headline variations (under 8 words)
– 5 first-line variations (the opening of the ad body)
– Each pair maps to a different customer pain or objection
– For each, tell me which audience segment it targets and which platform/placement fits
– Mark the one most likely to get the highest CTR and the one most likely to get the highest conversion-from-click (these are different)
Why this works: Ten variations costs nothing in the AI generation step and saves substantial time in the human review and platform-testing step. Asking AI to mark the click vs. convert tradeoff teaches the marketer the variation strategy without requiring 100 ad spend dollars to learn it the hard way.
8. Content repurposing
Repurpose it into:
– 3 short-form social posts (one for LinkedIn, one for X, one for Threads), each anchored on a different key insight from the source
– 1 email newsletter blurb (120 words) that links back to the source
– 1 short video script (30-45 seconds) for a TikTok or Reel
– 1 carousel outline (5-6 slides) for Instagram/LinkedIn
For each, name the specific source paragraph it came from. Do not introduce information not in the source.
Why this works: Most AI repurposing prompts produce derivative work that drifts from the original. The “do not introduce information not in the source” + “name the specific source paragraph” combination keeps the repurposed work tied to the asset. The result is a coordinated distribution pack that says the same thing five different ways without contradicting the original.
What is the worst thing you can do with AI in marketing?
- Skip prompts 1-3. Strategy, voice, and audience research are the inputs that make prompts 4-8 useful. Without them, you generate the average internet landing page, the average internet email sequence, the average internet social post. The “average internet” is what AI defaults to.
- Use AI for thought-leadership without a real point of view. AI is a poor source of original opinion. It is an excellent assistant to a human with a point of view; let it draft, but never let it originate.
- Fabricate metrics in case studies. “Increased ROI by 312%” without provenance is the line that gets quoted back at you in a deposition. Every metric needs a source you can defend.
- Run AI-generated ad creative without flagging it to the platform. Meta and TikTok both require AI-disclosure for synthetic-media ads in 2026; the consequence of non-disclosure has been account-level enforcement, not just ad-level rejection.
- Generate a year of social posts in one batch. The volume looks productive; the result is content that all reads the same, all gets flagged together, all underperforms. Use the prompts on a per-campaign basis, not a per-year basis.
What if your team writes 50+ marketing assets a month?
If you are running marketing at volume, the eight prompts above are a candidate for the next rung of the ladder. Save each one as a Claude skill with the voice brief and audience research baked in. Bundle them as a plugin called brand-marketing-stack that you invoke with one command for any campaign brief. The setup is a half-day; the time saved per month is 20-30 hours. After three months, the system has paid for itself many times over and the team has a portable artifact that survives turnover.
📊 The Prompt-to-Workflow Ladder
Tier 1: the prompts (this post). Tier 2: the skill (one per marketing function). Tier 3: the plugin (the brand-marketing-stack bundle). Tier 4: the workflow (the stack runs end-to-end on a brief and produces a coordinated drop). When to climb →
What are common questions about AI for marketing?
Should we disclose AI use in our marketing?
On paid ad placements, yes (Meta and TikTok require AI-disclosure for synthetic-media ads as of 2026; LinkedIn is moving the same direction). On organic content, no broad mandate, but the closer the asset is to a personal communication (a founder LinkedIn post, a sales email), the higher the cost of being caught and the lower the cost of being transparent.
Will LinkedIn down-rank AI marketing posts?
Yes for content that pattern-matches as AI, less for content that has been edited to remove the patterns. The behavior described by LinkedIn-marketing communities in late 2025 was soft suppression in the feed without any notification to the poster. The prompt-6 self-flag step is the practical mitigation.
How do we keep AI from killing our brand voice?
Prompt 2 (brand voice extraction) is the answer. A 1-page voice brief used as a system prompt for every subsequent generation. Without the voice brief, AI defaults to the global average; with it, AI defaults to your average.
What about Google AI Overviews killing our top-of-funnel?
Plan for it. Top-of-funnel pages that compete for summarized queries will see CTR declines; the strategy is to either (a) write for the summarization layer (so AI Overviews quote you when answering the query), or (b) shift effort to mid-funnel and bottom-funnel pages where the click is intent-loaded. Both are legitimate; the wrong move is pretending Overviews are not happening.
Where do these prompts come from?
They are the marketing section of the larger AI Prompt Library. The Library has over 500 prompts across 33+ categories with three difficulty levels, including expanded marketing coverage: SEO research, podcast guesting, partnership pitches, PR outreach, and conversion-rate optimization.
Sources to read next?
- HubSpot marketing blog
- Marketing Week
- Harvard Business Review marketing research
- Meta policy updates (AI disclosure rules)
- Search Engine Optimization (Grokipedia)
✏️ Before any asset ships
Brand-voice drift and AI pattern-matching together are what turn productive AI marketing into noise. The full 29-pattern catalog (Wikipedia's “Signs of AI writing” + the open-source Humanizer skill from Siqi Chen, MIT license) is in our cornerstone guide. Marketing teams running AI at volume should standardize on the skill as a pre-publish step in the brand-marketing-stack workflow. Our pre-publish editing pass: How to Edit AI Out of Your Writing →
The AI Prompt Library · $39
over 500 tested prompts including the full marketing section
The eight marketing prompts above are a free preview. The full Library has SEO research prompts, podcast guesting templates, partnership pitch scripts, PR outreach with embargo handling, CRO research templates, and the harder situations: rebrands, crisis comms, and category-creation campaigns.
1-on-1 Deep Work Session with James · $175
Build your brand-marketing-stack in 2 hours
A focused 2-hour session. We do prompts 1-3 live on your business (campaign strategy, voice extraction, audience synthesis), then we save them as skills and bundle the brand-marketing-stack plugin so your team can ship a coordinated drop in 90 minutes.
Get Smarter About AI Every Morning
Free daily newsletter. Built for people who want to use AI well, not chase every model.
Free forever. Unsubscribe anytime.
You may also like
- Best AI Prompts for Blog Writing — long-form that earns the click
- Best Claude Prompts for Sales — cold outreach + objection scripts
- Best AI Prompts for LinkedIn — the surface that fills your inbox
- How to Edit AI Out of Your Writing — strip AI patterns before you publish
- Prompt to Workflow: The AI Ladder — when to climb from prompt to workflow