The Claude + Meta Ads Playbook

Most people run Meta ads on gut feel: keep what feels like it's working, kill what feels slow, guess at the next creative. This playbook replaces the gut feel with three prompts that turn Claude into an ads analyst, a creative machine, and a hook library, all in one chat.

What's Inside

The Audit Prompt: paste your Ads Manager data, get a kill / scale / iterate teardown

The Creative Generator: turn one winning ad into 30 variants for testing

The Viral Hook Swipe File: 10 hook structures with examples and when to use each

Before You Start

✓

A Claude account. Free works for the audit and hooks, Pro is worth it for the variant generator if you're testing across multiple ads.

✓

Your last 30 days of Meta Ads Manager data (CSV export or a screenshot of the campaign view).

✓

One ad that's currently working as your control. Skip to the Hook Swipe File if you don't have one yet.

1. The Audit Prompt

What it does: Takes your Meta ads data and returns a structured audit: what to kill, what to scale, what to iterate, what's working across the account, and what's missing entirely. Three concrete actions for the next 7 days.

How to use it: Paste the prompt below into Claude. After the prompt, paste your campaign data. Claude returns the audit in about 30 seconds.

Where to get the data: Meta Ads Manager → select last 30 days → Export → CSV. Or screenshot the campaign view if there's not much going on.

Act as a senior performance marketer who has managed significant Meta ad spend. I'll paste my campaign data below. Return a structured audit covering: 1. KILL list. Which ads are draining budget without ROAS justification? Frequency too high? CTR too low? CPM out of line? Be specific about why each one goes. 2. SCALE list. Which ads are showing strong ROAS with room to grow? What's the next 7-day budget recommendation for each (percentage increase, not flat number)? 3. ITERATE list. Which ads have signal but aren't winning yet? For each, name the single biggest variable to test next: hook, creative format, audience, or offer. 4. Pattern read. What hooks, angles, and offers are working across the account? What's not? What's missing entirely from my testing matrix? 5. Next 7 days. Three concrete actions I should take this week, ranked by impact. Use these benchmarks as context (don't quote them in the output, just use them to judge): - Strong ROAS: 3x+ for ecom, 5x+ for low-ticket digital, breakeven OK for high-ticket if pipeline is healthy - Frequency: kill at 3.5+ unless retargeting - CTR floor: 1.0% on cold, 2.0% on warm - Strong CPC: under $1.50 cold, under $0.80 warm If data is missing for any of these, flag it and tell me what I need to pull next. Here's my data: [paste your Ads Manager export below]

💡 The KILL list is the most actionable part of the output. Most accounts have 30-50% of spend going to ads that should already be off. Cutting that first is often the single biggest ROAS lift you can make in a week.

2. The Creative Generator

What it does: Takes one ad that's working and generates 30 variants across 5 strategic categories: hook angle, format, length, CTA, and audience. You get a structured table you can paste straight into your testing spreadsheet.

How to use it: Paste the prompt into Claude, then paste your winning ad copy after it. Claude returns 30 variants organized by category.

Act as a senior performance marketer with deep Meta ad spend experience. Below is my winning ad. Generate 30 variant ad copies I can test, organized into 5 categories of 6 variants each: 1. Hook angle variants. Six different hook angles: curiosity, fear of missing out, contrarian take, social proof, specificity, direct benefit. 2. Format variants. Six different formats: story, listicle, question, testimonial, before-after, data-led. 3. Length variants. Six different lengths: one-liner, two-line, short paragraph, mini-story, long-form, multi-line scannable. 4. CTA variants. Six different CTAs: soft, hard, question, urgency, social-proof close, no-CTA. 5. Audience variants. Six rewrites for different audience temperatures: skeptics, beginners, advanced operators, mid-funnel, cold, retargeted. For every variant: - Keep the brand voice from the original (study it before drafting) - Match the same offer and value prop - Don't invent claims I haven't made - Don't use any of these clichés: "in today's world," "let's dive in," "the truth is," "unlock," "game-changer," "revolutionary," excessive exclamation points Output as a markdown table with columns: Variant #, Category, Sub-angle, Copy. Here's my winning ad: [paste your winning ad copy below]

💡 Don't test all 30 at once. You'll run out of statistical power and burn budget on noise. Pick 5-7 variants across categories, test in pairs against your control, kill losers in 3-5 days based on early signal, and double down on winners with a budget bump.

3. The Viral Hook Swipe File

10 hook structures, a real-feeling example for each, and when to use it.

The Specificity Hook

Structure: "I [specific result] using [specific tactic]"

Example: "I scaled a $5k/mo client to $50k/mo using one ad and three audiences"

When to use: Case studies, performance proof, building authority. Specificity beats generality every time. Replace "a lot" with the actual number.

The Contrarian Hook

Structure: "Stop [common advice]. Do this instead."

Example: "Stop A/B testing your creative. Do this instead."

When to use: Reaching experienced operators, breaking through ad fatigue. Works because it implies you know something the competition doesn't.

The Pattern Interrupt

Structure: "If you're [specific situation] and aren't [specific tactic], you're leaving money on the table."

Example: "If you're spending $5k+/mo on Meta ads and aren't using AI to generate variants weekly, you're leaving money on the table."

When to use: Pre-qualifying high-spend buyers. The qualifier filters out tire-kickers and stops the right people scrolling.

The Question Hook

Structure: "Why does [counterintuitive thing] keep happening?"

Example: "Why do my best ads always die after 2 weeks?"

When to use: Nurturing problem-aware audiences. Opens a curiosity gap that demands the next line.

The Cost-of-Inaction Hook

Structure: "Your competitors are [doing X]. You're [doing less]."

Example: "Your competitors are running 30 ad variants this week. You're running 3."

When to use: Competitive niches, urgency creation. Comparison is more visceral than an abstract benefit.

The Bold Promise Hook

Structure: "Every [thing] should follow this [number]-step formula."

Example: "Every Meta ad you write should follow this 3-line formula."

When to use: Educational content, lead magnets, building expert authority. Confident outcome statements work when you can back them up.

The Behind-the-Scenes Hook

Structure: "Here's the exact [thing] I [verb] this week."

Example: "Here's the exact $4,200 ad test I ran this week and what it told me."

When to use: Building trust and authority. Transparency creates connection. Real numbers beat vague claims.

The Mistake Hook

Structure: "I lost [specific amount] learning [lesson] so you don't have to."

Example: "I burned $8k learning this Meta targeting lesson so you don't have to."

When to use: Educational content. Pain plus benevolence is one of the most powerful framings in advertising.

The Stack Hook

Structure: "[Number] things every [winner] has + [number] things every [loser] has."

Example: "3 things every winning Meta ad has + 2 things every loser has."

When to use: Carousel, reel, or thread formats. The paired structure is scannable and creates a built-in hook for the second half.

The Industry Insider Hook

Structure: "What every [authority figure] knows that [non-experts] don't."

Example: "What every $10k+/mo Meta advertiser knows that beginners don't."

When to use: Positioning content, premium offer pre-sell. Implies exclusive knowledge and signals you're inside the room.

The honest bit: These prompts turn data you already have into a faster decision. They don't replace having a real offer or real ad spend to learn from. Claude reads the numbers you paste in. It doesn't pull live Ads Manager data on its own. Run the audit weekly, not once, and let the KILL list actually change your budget.

Want more of this? The Free 14-Day AI Challenge sends one build like this to your inbox every day.

Join the 14-Day Challenge →