Prompt · Marketing

Ad copy variants for A/B testing

Generate ad headlines and descriptions that each test one clear idea, such as a benefit, a pain point or social proof, so your A/B test tells you why one version won, not just which.

The prompt

You are a performance marketer. Write ad copy variants designed for a clean A/B test.

Product or offer: [WHAT YOU SELL AND THE SPECIFIC OFFER]
Audience: [WHO WILL SEE THE AD]
Platform and placement: [e.g. "search ads", "Instagram feed", "LinkedIn sponsored post"]
Character limits: [e.g. "headline 30, description 90"]
Proof points I can use: [REAL NUMBERS, REVIEWS, AWARDS, GUARANTEES]
Words or claims to avoid: [e.g. "cheap", "best", regulated claims]
Landing page promise: [WHAT THE PAGE AFTER THE CLICK SAYS]

Create 5 variants. Each must test ONE angle:
A. Main benefit
B. Pain point or problem
C. Social proof (use only my proof points)
D. Urgency or a specific offer detail (only if the offer is real)
E. Curiosity or a question

For each variant give:
- Angle and the hypothesis in one sentence ("Readers who ... will click because ...")
- Headline and description within the character limits (show character counts)
- Call to action

Then recommend which two variants to test first and why, and what minimum result would make you call a winner. Keep every variant consistent with the landing page promise.

Example output

Example (excerpt):
B. Pain point
Hypothesis: Small shop owners who dread month-end bookkeeping will click on relief from that task.
Headline: Month-end books in 20 minutes (28)
Description: Connect your bank, approve the matches, done. Free for 30 days, no card needed. (82)
CTA: Start free trial

C. Social proof
Headline: Rated 4.7 by 1,200 shop owners (29)

Test first: A vs B, because they differ only in framing and share the same offer.

When to use it

Use this prompt when you are about to launch or refresh a paid campaign on search, social or display. It helps whenever you have a budget for testing but tend to write five versions that are all slight rewordings of the same idea. Tests like that rarely teach you anything; tests built around different angles do. The same approach works for email subject lines and landing page headlines.

How to customize it

[PRODUCT OR OFFER] should be specific: “20% off the first order this week” tests better than “great prices”.
[AUDIENCE] shapes which angle is likely to work. Beginners respond to reassurance; experts to specifics.
[PLATFORM] and [CHARACTER LIMITS] should come from the ad platform’s current documentation, since limits change.
[PROOF POINTS] must be real and verifiable. Reviews, ratings and numbers you cannot back up can break advertising rules.
[WORDS TO AVOID] protects you from regulated claims in sectors such as health or finance.
[LANDING PAGE PROMISE] keeps ad and page consistent, which matters for both trust and ad quality scores.

Why it works

A good A/B test changes one variable at a time. Assigning each variant a single angle, with a written hypothesis, means the result tells you something about your audience you can reuse in emails, pages and future ads.

Showing character counts helps you catch overruns, though models can miscount characters because they process text as tokens rather than letters, so check in the ad editor. Restricting social proof and urgency to real information avoids invented claims. Asking for a minimum result before you start is a guard against declaring a winner too early from a small sample.

Tips and variations

Before launching, read each variant as a sceptical customer would and remove any claim you would not be comfortable defending in public or to the ad platform’s review team.
Ask for a version translated and adapted for another market, not just translated.
Add “Write two more variants of the winning angle” after your first test ends.
If your platform can rotate many headlines automatically, still group them by angle so you can read the report.
Our overview of how AI is changing marketing and advertising covers the wider trends.

#Ad Copy #A/B Testing #Marketing #Copywriting

More Marketing prompts

View all

Marketingintermediate

Brand voice guide from your best writing

You are a brand strategist and editor. Create a brand voice guide based on real examples. Brand and what it does: [ONE OR TWO SENTENCES] Audience: [WHO WE WRITE FOR] Three words we want to be described as: [e.g. "expert, warm, straightforward"] Words we never want to be described as: [e.g. "corporate, pushy, cute"] Samples that sound like us at our best (label each): """ [PASTE 3-6 SAMPLES: WEBSITE COPY, EMAILS, SOCIAL POSTS] """ One sample that does NOT sound like us, if available: """ [PASTE OR WRITE "none"] """ Produce: 1. Three to four voice traits. For each: a definition, what it is not, and a short example from my samples. 2. A "do / don't" table with paired example sentences. 3. How the voice adapts by channel: website, email, social media, error messages or support replies. 4. A word list: words we use, words we avoid, and preferred spellings or terms. 5. Sentence-level rules: length, use of "we" and "you", contractions, punctuation, emoji. 6. Two rewrites of the off-brand sample, or of a generic sentence, showing the voice in practice. Rules: base every trait on patterns in my samples and quote them. If the samples contradict each other, point that out instead of hiding it.

Works with: ChatGPT, Claude, Gemini, Any LLM

Marketingbeginner

Customer case study from an interview transcript

You are a B2B content writer. Write a customer case study from the interview below. Our product or service: [WHAT WE PROVIDE] Customer: [COMPANY, INDUSTRY, SIZE] and interviewee: [NAME, ROLE] Audience for the case study: [e.g. "operations leads at similar companies"] Target length: [e.g. "600-800 words"] Interview transcript or notes: """ [PASTE TRANSCRIPT] """ Produce: 1. Three headline options that state the result, not the product. 2. A summary box: customer, challenge, solution, results (bullet points). 3. The story in four parts: the situation before, why they looked for a solution, how they implemented it, and the results. 4. Two or three direct quotes taken word for word from the transcript, lightly cleaned of filler words only. 5. A list of every number and claim in the draft, with the line in the transcript it comes from. 6. Questions to send the customer to fill gaps or confirm figures. 7. A 60-word version for social media and a one-line version for a sales deck. Rules: do not invent results, numbers or quotes. If the transcript has no measurable result, describe the qualitative change and suggest what to measure. Keep any criticism of our product the customer mentioned in the notes for me, not in the case study.

Works with: ChatGPT, Claude, Gemini, Any LLM

Marketingintermediate

Customer persona built from your real research

You are a customer research analyst. Build a customer persona from the evidence below. Do not invent details that are not supported by it. Product or service: [WHAT YOU OFFER] Segment I am focusing on: [e.g. "first-time buyers", "small agencies"] How I will use the persona: [e.g. "to write website copy", "to prioritise features"] Evidence (label each source): """ [S1] [INTERVIEW NOTES] [S2] [CUSTOMER REVIEWS] [S3] [SUPPORT TICKETS OR SURVEY ANSWERS] """ Produce the persona with these sections: 1. Name and one-line description (make clear the name is fictional). 2. Context: role, situation and what triggers them to look for a solution. 3. Goals and jobs to be done. 4. Pains and frustrations, using their own words where possible. 5. Objections and doubts before buying. 6. Where they look for information and whom they trust. 7. Words they use (and words they never use). Rules: - After every point, cite the source labels it comes from, e.g. [S1, S3]. - Mark anything that is a reasonable inference but not directly stated as "ASSUMPTION". - Do not add demographics such as age, income or gender unless the evidence states them. - End with "Evidence gaps": 5 questions to ask in the next customer interviews.

Works with: ChatGPT, Claude, Gemini, Any LLM