When to use it
Use this when launching or refreshing a paid campaign and you need a batch of genuinely different variations to test, not five rewordings of the same sentence. It earns its keep when you have a single clear offer and a defined persona to organize the angles around, and a real testing budget where learning which ANGLE works is worth more than the copy itself. Works across LinkedIn, Google, Meta, and X if you specify the platform and its limits.
Do NOT use it as a creative crutch with a fuzzy offer and a vague audience, it will produce polished copy for a campaign that shouldn't run. And do NOT ship the variations that 'sound great' without a test structure; the point isn't to find one clever line, it's to learn which angle resonates so the next ten campaigns get cheaper. Never let it invent a stat or a customer to make a 'proof' angle stronger, a claim you can't back is a liability, not a hook.
The principle: paid testing is for learning angles, not admiring wordsmithing. The most common waste is testing five variations that are secretly the same message, learning nothing, and crowning the one that happened to win the coin flip. This prompt forces structurally different angles (pain, outcome, proof, contrarian, direct) so a test actually tells you something you can reuse.
The prompt
Prompt, paste into Claude, ChatGPT, Gemini
You are a direct-response copywriter who writes paid ads that convert. You write distinct ANGLES, not synonyms, if two variations could be A/B tested and teach the same lesson, you've failed. Every line is specific and benefit-led. You never use hype words and never invent stats, customers, or claims I can't back.
CONTEXT
Platform + ad format: {{PLATFORM}}
Product / offer: {{OFFER}}
Target audience: {{AUDIENCE}}
Core pain: {{CORE_PAIN}}
Key benefit / outcome: {{KEY_BENEFIT}}
Proof or differentiator (optional, must be defensible): {{PROOF}}
Call to action: {{CTA}}
Constraints (character limits, brand do's/don'ts): {{CONSTRAINTS}}
TASK
Write 5 ad variations, each using a STRUCTURALLY different angle so a test of them teaches us which message works.
METHOD
1. Use these five angles, one per variation: (1) Pain-led, (2) Outcome/aspiration-led, (3) Social-proof/credibility-led, (4) Contrarian/pattern-interrupt, (5) Direct/no-nonsense.
2. For each, write the strongest possible version of THAT angle, don't blend angles or hedge.
3. Respect the character limits in {{CONSTRAINTS}}; report the character count for each headline.
4. If PROOF is thin, lean on specificity and concreteness rather than manufacturing a claim.
OUTPUT FORMAT (per variation)
[Angle name]
Headline: (char count)
Primary text / body:
CTA:
Then add:
TEST FIRST: which angle to test first and the reasoning (usually the one that also qualifies the click).
VISUAL IDEAS: one image/visual concept for each of the top two angles.
KILL CRITERIA: one line on what result would tell you an angle is a loser worth cutting.
CONSTRAINTS
- Banned words: unlock, leverage, supercharge, seamless, game-changer, revolutionize, effortless, best-in-class, world-class.
- No invented stats, customers, or claims. If proof is weak, say so and compensate with specificity.
- Honor every character limit; flag any headline you couldn't fit and offer a shorter alternative.
- If OFFER or AUDIENCE is vague, tell me the variations will be generic and ask one sharpening question before writing.Run it from the terminal
$# key is in the env, never hardcoded
$echo ${OPENAI_API_KEY:+set}
set
$# generate 5 distinct ad angles with ChatGPT (gpt-4o)
$cat prompts/ad-copy.md | llm -m gpt-4o
1. Pain-led
Headline: Spending $20k/mo on LinkedIn and can't trace the pipeline? (58)
Body: If spend is up but pipeline is murky, the leak is usually in three places.
We'll find them in a free audit, no pitch.
CTA: Get a free audit
2. Outcome-led
Headline: Same budget, cleaner pipeline from LinkedIn (44)
3. Social-proof-led
Headline: We've audited 50+ B2B LinkedIn accounts (40)
4. Contrarian
Headline: Most LinkedIn ad problems aren't a budget problem (52)
5. Direct
Headline: Free LinkedIn Ads audit. No pitch. (33)
TEST FIRST: Pain-led, it qualifies on spend ($20k/mo) so clicks self-select for fit.
VISUAL IDEAS: Pain-led, a funnel with a visible leak. Social-proof, a '50+' stat card.
KILL CRITERIA: CTR below half the leader's after ~5k impressions with higher CPL, cut it.
$# five structurally different angles, every headline under 70 chars
$
Variables to swap
| Variable | What to put there |
|---|---|
{{PLATFORM}} | Where the ad runs and the format (LinkedIn single image, Google search, Meta, X) |
{{OFFER}} | The product or specific offer being promoted |
{{AUDIENCE}} | Who you're targeting |
{{CORE_PAIN}} | The problem your audience feels |
{{KEY_BENEFIT}} | The main outcome you deliver |
{{PROOF}} | A defensible result, customer, or differentiator (optional) |
{{CTA}} | The action you want (Book a demo, Get the guide) |
{{CONSTRAINTS}} | Character limits and brand do's and don'ts |
Worked example
Input
Filled-in example
PLATFORM: LinkedIn (single image ad)
OFFER: Free LinkedIn Ads audit
AUDIENCE: B2B demand gen managers spending $20k+/mo on LinkedIn
CORE_PAIN: Spend is high, pipeline from it is unclear
KEY_BENEFIT: Find the budget leaks and lower CPL
PROOF: 50+ B2B accounts audited
CTA: Get a free audit
CONSTRAINTS: Headline under 70 chars; no superlatives like 'best'Output
What the model returns
1. Pain-led
Headline: Spending $20k/mo on LinkedIn and can't trace the pipeline? (58)
Body: If your LinkedIn spend is up but the pipeline is murky, the leak is usually in three places. We'll find them in a free audit, no pitch.
CTA: Get a free audit
2. Outcome-led
Headline: Same budget, cleaner pipeline from LinkedIn (44)
Body: Lower your CPL without raising spend. Start with a free audit of your current targeting, structure, and creative.
CTA: Get a free audit
3. Social-proof-led
Headline: We've audited 50+ B2B LinkedIn accounts (40)
Body: The same budget leaks show up again and again. Find yours in a free, no-pitch audit.
CTA: Get a free audit
4. Contrarian
Headline: Most LinkedIn ad problems aren't a budget problem (52)
Body: They're a targeting and structure problem. A free audit shows you where the money actually goes.
CTA: Get a free audit
5. Direct
Headline: Free LinkedIn Ads audit. No pitch. (33)
Body: Send your account; get a clear list of what's leaking budget and how to fix it.
CTA: Get a free audit
TEST FIRST: Pain-led, it qualifies on spend ($20k/mo) and names the exact frustration, so the clicks self-select for fit.
VISUAL IDEAS: Pain-led, a funnel with a visible leak/drip. Social-proof, a clean '50+ accounts audited' stat card.
KILL CRITERIA: If an angle's CTR is below half the leader's after ~5k impressions and CPL is higher, cut it and reallocate.Tips to get more out of it
- Force distinct angles or you learn nothing. If two variations feel similar, tell it 'make these meaningfully different angles, not reworded versions of the same message.'
- Always pass real character limits in CONSTRAINTS, LinkedIn, Google, and Meta all truncate differently, and a headline that gets cut mid-word kills the variation regardless of how good it was.
- Test ONE variable at a time. Run the five angles first to find the winning message, then test headline variants within the winning angle, don't change angle and image and CTA at once or the data is noise.
- Honor the KILL CRITERIA. The discipline of cutting losers fast is what makes the budget compound; teams that keep 'pretty' losers alive bleed CPL.
- Keep the specificity-led variants over claim-led ones when proof is thin. A concrete '$20k/mo' beats a vague superlative every time, and it won't get you a fact-check.
- Pull the winning ad's hook into your organic posts and email subject lines, paid is a fast, paid way to learn which message resonates; reuse the answer everywhere.
- Across models: ChatGPT and Claude both do well here; Gemini sometimes ignores character counts, so always verify counts yourself before uploading.