Resolve bracketed inputs and the control ad from the current conversation, active project, uploaded files, and available connected sources before asking. Do not ask the user to repeat confirmed details.
Design [NUMBER OF TESTS] clean A/B tests from this control ad:
[CONTROL AD]
Context:
Offer: [OFFER]
Audience: [AUDIENCE]
Platform: [PLATFORM]
Objective: [OBJECTIVE]
Approved facts: [APPROVED FACTS]
Available headline formulas: Benefit in a Box - [Clear benefit] for [target audience]; Pain Point Flip - Tired of [frustration]? Try [solution].; Outcome + Timeframe - Get [desirable result] in [time]; Social Proof Snapshot - [Number or group] already [positive result]; Intriguing Question - What if you could [solve problem] without [common pain]?
Available body formulas: Problem → Solution → Outcome; Rapid Benefit List - [Product] helps you [benefit], [benefit], and [benefit]-without [usual hassle].; Short Testimonial - I used to [struggle], but now I [win]. - [Persona/Title]
Available CTA formulas: Value + Action - [Benefit] → [Action]; First-Person Empowerment - [Action] my [benefit]; Curiosity-Based CTA - Want to [solve problem]? → [Action]; Instructional Prompt - Click to [outcome]; Soft Awareness CTA - Explore [topic/resource]
Change one major variable per test while keeping the offer, audience, proof, placement, and action constant. A formula may be the tested variable, but do not change another strategic element in the same pair. Do not use proof-dependent formulas without approved evidence.
For each test, return: hypothesis, variable changed, Version A with formula labels, Version B with formula labels, elements held constant, primary metric, and interpretation caution.