Write 10 Ad Variations That Each Test One Variable Against a Control
You get a control ad plus 10 single-change variants (hook, offer framing or CTA), each with a labeled hypothesis and primary metric so A/B test results are easy to read.
Prompt
You are a performance marketer who designs clean A/B tests for paid ads.
Write 10 ad variations that each change exactly one element of a control ad, so results are easy to read.
First, ask me these in one message and wait:
1. What are you selling, and what is the offer?
2. Who is the audience and what problem does it solve?
3. Which platform, and what character limits apply?
4. Paste your current ad as the control, or say "none" and write one first.
Rules:
- Each variant changes only the hook (first line), the offer framing, or the CTA. All other text stays word for word.
- Split: 4 hook, 3 offer framing, 3 CTA variants.
- Each variant gets a hypothesis: "Changing X to Y will raise Z because..."
- Name one primary metric per variable type.
- Respect character limits. Do not add claims, numbers or testimonials I did not give you.
Output format:
1. Control ad (headline, primary text, CTA)
2. Table: ID | Variable | New text | Hypothesis | Primary metric
3. Three bullets on test order and budget per cell.
End with one line on assumptions or risks, such as too little traffic to test all 10 at once.Example output
Tips and common edits
Why this works
- Locking all text except one element means any lift can be traced to a single change.
- A written hypothesis per variant makes it clear what you learned, even when a variant loses.
- Running hooks first, then offer, then CTA builds a stronger control each round.
Use before launching a new Meta, Google or LinkedIn ad test. Have your offer, audience, platform limits and current best ad ready. Works in ChatGPT, Claude or Gemini.
Common edits
- Change the split to 10 hook variants if you only want to test openers this round.
- Add a fourth variable type, such as image description or headline, with its own metric.
Prompt details
- By
- Prompts-GPT editors
- Published
Did this prompt work?
Keep exploring
More like this
Turn Customer Pain Points into 20 Ad Hooks Ranked by Specificity
You get 20 short ad hooks split into question, statement and scenario types, each tied to a real pain point and ranked by how concrete it is.
Write Facebook and Instagram Ad Copy in Pain, Outcome and Curiosity Angles
Get three ready-to-test Meta ad variations, each with primary text hooked in the first 125 characters, a headline, a description and a CTA button.
Audit an Underperforming Ad and Rewrite It With a Reason for Every Change
Get a pass/fail audit of your pasted ad on four copy checks, two rewrites that test different angles, and a change log tying every edit to a failed check.
Write a LinkedIn Sponsored Content Ad That Speaks to One B2B Job Title
You get three LinkedIn single-image ad variations aimed at one job title's problem, with intros under 150 characters, headlines, descriptions, CTAs and a test pick.
Write a YouTube Pre-Roll Ad Script That Lands Before the Skip Button
You get a timed YouTube pre-roll script that delivers your brand and core promise in the first five seconds, plus a 6-second bumper cut and two alternative hooks.
Write and Rank 15 Email Subject Lines with Preview Text for A/B Testing
You get 15 subject lines under 50 characters, each paired with preview text that adds a new fact, tagged by type, ranked, plus two ready-to-run A/B test pairs.