Prompt Basics for Marketing: Clear Instructions, Better Outputs

1) A Prompt Formula Tailored to Marketers

Marketing prompts work best when they behave like a tight creative brief: specific goal, clear audience context, and explicit output requirements. The model is very good at generating options; your job is to constrain it so the options are usable.

The marketer prompt formula

Use this order (you can copy/paste it as a checklist):

• Goal: What you need and why (e.g., “increase demo sign-ups,” “reduce churn,” “drive webinar registrations”).
• Audience: Who it’s for, including segment and stage (cold/warm/hot; awareness/consideration/decision).
• Channel: Where it will appear (email, LinkedIn post, landing page hero, paid search ad, SMS, etc.).
• Tone / brand constraints: Voice, reading level, do/don’t words, compliance notes.
• Key points: Must-include facts, differentiators, proof points, offer details.
• Format: Structure (bullets, AIDA, PAS, table, JSON, headings, etc.).
• Length: Word/character limits, number of options.
• CTA: Exact call-to-action and link placeholder.
• Exclusions: What to avoid (claims, competitors, sensitive topics, certain adjectives, emojis, etc.).

Fill-in prompt skeleton

Goal: [what you want the audience to do/think/feel]  
Audience: [segment + stage + pain points + objections]  
Channel: [where this will be used + any platform constraints]  
Tone/Brand constraints: [voice, reading level, do/don’t, compliance]  
Key points to include: [bullets of facts, differentiators, proof]  
Format: [required structure: bullets/table/headings/etc.]  
Length: [word/character count + number of variants]  
CTA: [exact CTA text + link placeholder]  
Exclusions: [what to avoid]  
Output: [exact deliverable instructions]

Example: turning a vague request into a usable brief

Vague prompt:

Write a LinkedIn post about our new analytics feature.

Brief-style prompt:

Goal: Drive clicks to a product update page and increase free-trial sign-ups.
Audience: B2B SaaS marketing managers at 50–500 employee companies; stage = consideration (they know they need better attribution, comparing tools).
Channel: LinkedIn organic post.
Tone/Brand constraints: Clear, confident, practical; no hype; 8th–10th grade reading level; avoid buzzwords like “game-changing” and “revolutionary.”
Key points to include: (1) New multi-touch attribution view, (2) faster dashboard load time (30% improvement), (3) export to CSV, (4) works with Google Ads + Meta.
Format: Hook (1–2 lines) + 3 bullet benefits + short proof line + CTA.
Length: 120–160 words.
CTA: “See what’s new” + [link].
Exclusions: No competitor mentions; no unverified ROI claims.
Output: Provide 3 post variants.

2) How Small Changes Affect Results

Small prompt edits often produce big quality jumps because they reduce ambiguity. The three highest-leverage additions for marketers are: audience stage, differentiators, and constraints.

Continue in our app.

You can listen to the audiobook with the screen off, receive a free certificate for this course, and also have access to 5,000 other free online courses.

Or continue reading below...

Download the app

Demonstration: one prompt, three small upgrades

Scenario: You want email copy for a webinar invitation.

Version A: minimal

Write an email inviting people to our webinar about email marketing.

Typical outcome: generic subject line, vague benefits, unclear audience, weak CTA.

Version B: add audience stage + objections

Write an email inviting people to our webinar about email marketing.
Audience: Ecommerce founders who are warm leads (downloaded our guide last week) but haven’t booked a call.
Objections: “I don’t have time” and “I’ve tried email before; it didn’t work.”

What improves: the email can acknowledge time constraints, reference prior engagement, and focus on practical outcomes.

Version C: add differentiators + constraints + structure

Goal: Get webinar registrations.
Audience: Ecommerce founders; stage = warm lead; objections = time + skepticism.
Offer: 45-minute live session + 15-minute Q&A.
Differentiators: Real examples from 3 stores; includes a swipe file; no advanced tools required.
Channel: Email.
Tone: Direct, helpful, non-salesy.
Constraints: No emojis; avoid phrases “unlock,” “secret,” “crush it.”
Format: Subject line + preview text + body with (1) problem, (2) what they’ll learn (3 bullets), (3) who it’s for (1 sentence), (4) logistics, (5) CTA button text.
Length: 140–180 words.
CTA: “Save my seat” + [registration link].
Output: Provide 5 subject lines and 1 full email.

What improves: you get a deliverable that fits your brand, includes specific proof points, and is ready to paste into your ESP with minimal editing.

Quick checklist: common “small changes” that raise quality

• Add stage: “cold awareness” vs “decision-stage comparing vendors” changes messaging dramatically.
• Add differentiators: “30% faster,” “done-for-you templates,” “SOC 2,” “2-week setup.”
• Add constraints: banned words, compliance rules, reading level, character limits.
• Add required structure: tables, numbered sections, specific fields.
• Add exclusions: no competitor names, no medical/financial promises, no absolute claims (“guaranteed”).

3) Reusable Prompt Templates (Copy/Paste)

These templates are designed to be reused across campaigns. Replace bracketed text with your details.

Template A: Summarize (for briefs, meeting notes, research)

Goal: Summarize the content so a marketer can act on it.
Audience: [role], stage = [awareness/consideration/decision].
Source text: [paste text]
Key points to extract: [e.g., pain points, benefits, proof, objections, next steps].
Format: 
- 5-bullet executive summary
- 3 key takeaways
- 3 recommended actions for our campaign
Length: [e.g., 150–250 words].
Exclusions: Do not invent facts; if missing, write “Not stated.”

Template B: Generate variants (headlines, hooks, CTAs)

Goal: Generate multiple on-brand variants for testing.
Audience: [segment + stage].
Channel: [e.g., LinkedIn ad / landing page hero / email subject line].
Product/offer: [what it is].
Differentiators: [3–5 bullets].
Tone/Brand constraints: [voice + banned words].
Format: Provide [N] variants. Each variant must:
- Mention [required term] (if applicable)
- Stay under [character/word limit]
CTA: [exact CTA or CTA style].
Exclusions: [no claims, no competitor mentions, etc.]
Output: Return as a table with columns: Variant #, Copy, Angle (benefit/proof/urgency), Notes.

Template C: Outline (blog post, landing page, webinar, nurture sequence)

Goal: Create an outline that supports [primary conversion goal].
Audience: [segment + stage + top 2 objections].
Channel: [blog/landing page/webinar/email sequence].
Key points to include: [bullets].
Format: 
- H2/H3 outline
- For each section: 1–2 sentence summary + suggested proof/examples
Length: [e.g., 900–1200 words final; outline only].
Exclusions: No fluff; avoid repeating the same point in multiple sections.

Template D: Rewrite (match tone, improve clarity, adapt to channel)

Goal: Rewrite the text for [channel] to improve clarity and conversion.
Audience: [segment + stage].
Tone/Brand constraints: [voice, reading level, do/don’t].
Original text: [paste]
Requirements:
- Preserve meaning and factual claims
- Improve scannability (short paragraphs, bullets where helpful)
- Add a clear CTA: [CTA text] + [link]
Length: [target word count or character limit].
Exclusions: Do not add new features, stats, or guarantees.

Template E: Extract insights into tables (voice-of-customer, reviews, interviews)

Goal: Turn the text into structured marketing insights.
Audience: Internal marketing team.
Source text: [paste reviews/interview notes/support tickets]
Task:
1) Identify recurring themes.
2) Extract exact phrases customers use.
3) Map themes to messaging angles.
Format: Output a table with columns:
- Theme
- Customer quote (verbatim)
- Underlying job-to-be-done
- Objection/risk
- Suggested message angle
- Suggested proof (what we’d need to claim it)
Rules: Use only evidence from the text; if proof is missing, mark as “Need proof.”

4) Mini-Lab: Iteratively Refine One Prompt (3 Rounds)

This mini-lab trains you to improve outputs by adding constraints and a required structure. You’ll start with a basic prompt and refine it in three rounds. Use any product/offer you’re currently marketing.

Setup (choose your scenario)

• Offer: [e.g., free trial, webinar, consultation, new feature]
• Audience: [segment]
• Channel: [e.g., LinkedIn post, email, landing page hero]
• One differentiator: [e.g., “setup in 15 minutes”]

Round 1: Basic deliverable

Your prompt (fill in the blanks):

Write [channel] copy to promote [offer].
Audience: [segment].
CTA: [CTA text] + [link].
Output: Provide 5 headline options.

What to check:

• Are the headlines specific to your offer, or generic?
• Do they match the channel (e.g., ad-like vs editorial)?
• Do they imply a clear benefit?

Round 2: Add stage + differentiators + exclusions

Now constrain the model so it chooses better angles.

Goal: Increase [primary metric].
Audience: [segment]; stage = [awareness/consideration/decision].
Top pain point: [pain].
Top objection: [objection].
Differentiators to include (choose at least 1 per headline):
- [diff 1]
- [diff 2]
Tone/Brand constraints: [voice]; avoid [banned words].
Exclusions: No guarantees, no competitor mentions, no “best/number one” claims.
Output: Provide 8 headline options.

What to check:

• Do the headlines address the stage (education vs comparison vs urgency)?
• Do they include differentiators instead of vague benefits?
• Are any headlines risky (overpromising)? If yes, tighten exclusions.

Round 3: Require output structure (table) + testing angles

In this round, you’ll force the output into a decision-ready format so you can pick winners quickly.

Goal: Produce headline options we can A/B test.
Audience: [segment]; stage = [stage].
Channel: [channel] with limit [X characters] per headline.
Differentiators: [list].
Tone: [brand voice].
Constraints:
- Must be under [X] characters
- No emojis
- No superlatives (best, #1, ultimate)
- Use active voice
Required angles: Create headlines across these angles (at least 2 each):
1) Pain-to-solution
2) Proof/metric
3) Speed/time-saving
4) Risk-reversal (without guarantees)
Output format: Return a table with columns:
- Angle
- Headline
- Character count
- Differentiator used
- Notes (why it might work)
Output: Provide 12 headlines total.

What to check:

• Does every row comply with character limits and constraints?
• Do you have a balanced set of angles for testing?
• Are differentiators distributed, or is the model repeating one?

Optional refinement moves (if results still feel off)

• Add examples of what “good” looks like: provide 2–3 sample headlines in your brand style and say “match this style.”
• Add a “must mention” field: e.g., “must mention ‘multi-touch attribution’ exactly once.”
• Add a “do not mention” list: e.g., avoid “AI,” avoid “automation,” avoid pricing.
• Add a scoring step: ask the model to rate each headline 1–5 for clarity and specificity, then revise the lowest-scoring ones.