Most Meta advertisers test too randomly. They launch a few creatives, pick the

best performer after a week, and call it a strategy.

The top buyers do something different. They test systematically — isolating

variables so every experiment tells them something the others can't.

That's the entire difference between a brand that finds a winning ad by luck

and one that builds a repeatable system for finding winners.

THE CORE INSIGHT
→ 50 random creatives give you noise
→ 27 structured experiments give you signal
→ The constraint was never ideas — it was the time between idea and live ad
→ Claude closes that gap entirely

3 variables. 3 options each. Every combination is a structured test with a clear

hypothesis. Here's the full system.

The 3 variables every Meta test needs to cover

Before you look at the prompt, understand the framework underneath it. Every

high-performing Meta ad is a combination of three decisions. Test them in

isolation and the data compounds. Test them randomly and it cancels out.

3 variables × 3 options = 27 structured experiments

Each combination is its own hypothesis. Fear hook + static image + low-friction

CTA tells you something completely different from aspiration hook + carousel +

high-intent CTA. That's the signal.

What the 3×3 matrix looks like in practice

Here's how the combinations map out across hook angle and format. Each cell

is one ready-to-brief experiment — hook type, format, and CTA defined before a

single word is written.

How Claude generates all 27 variants in under 10 minutes

Here's the exact sequence. One prompt, one paste, full swipe file output —

structured, labelled, and ready to brief your designer.

1 Reads your product and landing page

Claude takes your product URL or brief and extracts the core claims, benefits, and

emotional triggers. It reads the copy the way a cold prospect would — looking for

what would actually stop a scroll.

2 Maps every combination across the 3×3 grid

For each of the 9 format–hook combinations, Claude writes copy matched to that

specific angle. Not generic variations — each one is built around the emotional

logic of that cell.

3 Writes 3 CTA variants per combination

Each hook + format combination gets three CTA treatments: low friction ("See

how it works"), value exchange ("Get the free guide"), high intent ("Start your

trial"). That's where the 27 comes from.

4 Labels and formats the full swipe file

Every variant is tagged by hook type, format, and CTA — so you can filter by

angle and run structured tests rather than random experiments. Clean enough to

paste directly into a designer brief.

The prompt (copy this exactly)

Paste this into Claude with your product URL or brief. Replace only the words in

brackets — everything else is intentional.

PROMPT — PASTE INTO CLAUDE · REPLACE THE BRACKETS

# 3×3 Meta Creative Testing System
You are an expert direct-response Meta ads copywriter.
Your task: Generate 27 Meta ad copy variants for the
product below using a structured 3×3 testing framework.
Product URL / brief: [PASTE YOUR URL OR PRODUCT BRIEF]
Step 1 — Extract core emotional triggers
Read the product page and identify:
- The 3 deepest fears your customer has right now
- The 3 strongest aspirations they're chasing
- The 3 curiosity gaps your product can open
Step 2 — Generate copy across the 3×3 grid
For each combination of Hook angle × Format, write one
complete ad hook (max 15 words) + 3 CTA variants:
Variables:
- Hook angle: Fear / Aspiration / Curiosity
- Format: Static image / 15s video / Carousel
- CTA: Low friction / Value exchange / High intent
Step 3 — Format as a labelled swipe file
EXPERIMENT [number]: [Hook] × [Format] × [CTA type]
→ Hook copy: [first line, max 15 words]
→ CTA copy: [call to action text]
→ Brief note: [one line for the designer]
# Rules:
# - No em dashes, no AI phrases (unlock, leverage)
# - Each hook works as a standalone scroll-stopper
# - Fear hooks activate cost of inaction — not fear itself
# - Curiosity hooks open a loop without resolving it
# - Vary sentence length and rhythm across all 27
Product brief:
[PASTE HERE]

How to use the output

Getting 27 variants is step one. Here's how to turn them into actual test results

that compound.

THE TESTING FRAMEWORK
Run one combination per ad set, targeting the same cold audience. After 3–4
days and 1,000 impressions per variant, pause the bottom performers by CTR.
Reinvest into the top hook angle. Then test CTA variants within that winner. Layer
by layer until you have a clear champion per intent level

What you're building is a structured signal factory. Each test tells you not

just which ad performed — it tells you which variable drove the result. That's

data you can apply to every campaign you run from here.

Avoid these mistakes

The real advantage

The media buyers using this aren't replacing creative judgment. They're

offloading the volume problem to Claude so their judgment has more signal to

work with.

Your designer builds the master template once. You swap the hooks across

variants. No brief needed. No back-and-forth. Launch-ready in an hour.

The constraint was never ideas. It was the time between idea and live ad.

3 variables × 3 options = 27 experiments = real signal

More structured tests = cleaner signal = winning ads faster.

That's the whole game.