zmime.com
Category · Marketing

Marketing prompts

Marketing prompts get good when you feed them your data instead of asking for best practices.

  • SEO briefs
  • Paid ads
  • Social
  • Positioning
  • Launches
01 — Overview

Why these prompts look the way they do

Ask a model for "marketing ideas" and you get a listicle you could have written yourself. Ask it to work on your positioning, your customer language and your analytics numbers, and it becomes genuinely useful. The input material is the whole game.

Everything here assumes a human reviews claims before publishing. Models are confident about statistics, competitor features and compliance language, and confidently wrong often enough to matter.

02 — Anatomy

Inputs a marketing prompt needs

  • Positioning What you do, for whom, instead of what alternative.
  • Customer language Real quotes from reviews, support tickets or calls.
  • Objective + metric The single number this asset should move.
  • Channel + format Where it runs, and its length and character limits.
  • Proof Metrics, customers, integrations, guarantees you can substantiate.
  • Brand rules Voice, banned claims, legal and compliance constraints.
  • Review flags Ask the model to mark anything needing verification.
03 — The prompts

5 copy-ready marketing prompts

Replace everything in [brackets]. Keep the constraint lines — they do more work than the descriptive ones.

Nothing matches that word. Clear the field to see every prompt.

Prompt 01

SEO content brief from a keyword

Briefing a writer or generating a draft that matches intent.

  • seo
  • brief
Template · plain text
Create an SEO content brief for the primary keyword "[keyword]".

BUSINESS: [what you sell, to whom]
SEARCH INTENT (my read): [informational / commercial / transactional]
CURRENT TOP RESULTS (pasted below): [titles + a summary of what each covers]
OUR ADVANTAGE: [proprietary data, tooling, expertise, customer access]

DELIVER
1. Target intent in one sentence, and the searcher's actual question.
2. Recommended format and word count, with reasoning.
3. Title tag (max 60 chars) and meta description (max 155 chars).
4. H2/H3 outline where each H2 answers a distinct sub-question.
5. Entities and subtopics to cover (list, no keyword stuffing).
6. Questions to answer directly, sourced from the results below.
7. Content gaps: what the current results all miss that we can cover.
8. Internal link targets: [list our relevant pages].
9. What would make this page hard to copy.

COMPETITOR RESULTS:
"""
[paste]
"""
Prompt 02

Ad variations for structured testing

Paid social or search copy where you want to learn from the test.

  • ads
  • testing
Template · plain text
Write ad copy for [platform] promoting [product] to [audience].

OFFER: [what they get]
PAIN: [top pain in the customer's own words]
PROOF: [metric or customer]
LANDING PAGE PROMISE: [exact headline on the page — copy must match it]
CONSTRAINTS: headline max [n] chars, body max [n] chars, [platform] policy compliant, no unverifiable claims.

PRODUCE 5 VARIANTS, EACH TESTING ONE ANGLE
1. Pain-led
2. Outcome-led
3. Proof-led (numbers)
4. Objection-led
5. Curiosity-led (no clickbait)

For each variant give: headline, body, CTA, the hypothesis it tests, and the metric that would prove it.

Then flag any claim that needs verification before it runs.
Prompt 03

Social calendar from one source asset

Turning a single article or launch into a month of posts.

  • social
  • repurposing
Template · plain text
Turn the asset below into a [4]-week social calendar for [platforms].

AUDIENCE: [who], who cares about [topics]
GOAL: [awareness / signups / community], measured by [metric]
POSTING CADENCE: [n] posts per week per platform
VOICE: [description], plus these rules: [no emoji spam, no hashtag walls, no engagement bait]

FOR EACH POST GIVE
- Week and day
- Platform and format (text, carousel, short video, image)
- Hook (first line, under 12 words)
- Body, within the platform's practical length
- One idea per post — no summaries of the whole article
- Visual note (what to show)
- Link placement (in post / in comment / none)

RULES
- Every post must stand alone and deliver one specific insight.
- No claims beyond the source asset.
- Vary formats; do not repeat the same hook pattern more than twice.

SOURCE ASSET:
"""
[paste]
"""
Prompt 04

Positioning and messaging matrix

Getting a whole team to describe the product the same way.

  • positioning
  • messaging
Template · plain text
Build a messaging matrix for [product].

INPUTS
- What it does: [plain description]
- Segments: [segment A], [segment B], [segment C]
- For each segment, the job to be done: [...]
- Current alternatives: [competitor, spreadsheet, doing nothing]
- Proof: [metrics, customers, certifications]
- Customer quotes: [paste 5-10 real quotes from reviews or calls]

DELIVER A TABLE, ONE ROW PER SEGMENT
- Segment and trigger event
- Primary pain, in their words (draw from the quotes)
- Value proposition (one sentence, no jargon)
- Three supporting messages, each with a proof point
- Top objection and the response
- Words to use / words to avoid for this segment

THEN
- One positioning statement covering all segments.
- Where segments conflict, say so instead of averaging them.
- Mark any claim not supported by the inputs as [UNVERIFIED].
Prompt 05

Launch announcement kit

Shipping a feature or product across every channel at once.

  • launch
  • announcement
Template · plain text
Write the launch kit for [feature/product], shipping on [date].

WHAT CHANGED: [specific capability]
WHO IT HELPS AND HOW: [segment] can now [outcome] instead of [old workaround]
WHY NOW: [reason customers will find credible]
PROOF: [beta results, quotes, benchmarks]
NOT INCLUDED: [limitations to state honestly]

DELIVER
1. Blog post: 500 words — what, why, how to use it, limitations.
2. Email to existing users: 120 words, one CTA.
3. Changelog entry: 60 words, factual.
4. Social posts: one per [platform], each with a different angle.
5. In-app message: 200 characters.
6. Support FAQ: 5 questions real users will ask, with plain answers.
7. Sales one-liner and a 30-second talk track.

RULES
- Same core claim everywhere, adapted in length not in substance.
- State limitations in the blog post and FAQ, not only in docs.
- No "we're excited to announce" openers.
04 — Craft notes

What separates a good prompt from a wasted run

Do this 05

  • Paste real customer language from reviews and tickets.
  • Give one objective and one metric per asset.
  • Include character limits for every channel.
  • Ask for hypotheses alongside ad variants so tests are readable.
  • Require [UNVERIFIED] tags on any claim you did not supply.

Not this 05

  • Asking for "viral" content or generic best practices.
  • Publishing statistics the model produced on its own.
  • Letting output describe competitor features without checking.
  • One prompt for strategy, copy and design at once.
  • Skipping legal review on regulated claims.
05 — Questions

Asked often, answered plainly

How do I stop AI marketing copy from sounding like AI?

Remove the tells and add specifics: no "in today’s landscape" openers, no triads of adjectives, no hedging. Then insert facts only you have — numbers, customer quotes, product detail. Specificity is what reads as human.

Can I trust generated competitor comparisons?

No. Pricing and feature sets change constantly and models fill gaps confidently. Use the model to structure the comparison, then verify every cell against the competitor’s live pages.

How much should a human edit?

Assume you own every claim. Practically that means checking facts, cutting a third of the words, and rewriting the opening and closing lines, which are where generated copy is weakest.

06 — Sources

Read the primary documentation

Vendors change flags, limits and defaults often. Where this site disagrees with official docs, the docs win.