Analyzes your top performing ads, identifies what's working in the hooks, CTAs, messaging angles, and formats, then generates new variants that follow the same winning patterns while introducing enough variation to test meaningfully. Platform: Google and Meta.
git clone https://github.com/irinabuht12-oss/marketing-skills.git--- name: ad-copy-variant-generator description: Analyzes your top performing ads, identifies what's working in the hooks, CTAs, messaging angles, and formats, then generates new variants that follow the same winning patterns while introducing enough variation to test meaningfully. Platform: Google and Meta. metadata: platform: Google and Meta --- # 9/ Ad Copy Variant Generator — Google + Meta ## What it does Analyzes your top performing ads, identifies what's working in the hooks, CTAs, messaging angles, and formats, then generates new variants that follow the same winning patterns while introducing enough variation to test meaningfully. ## How it works Claude ingests your ad performance data alongside the actual ad copy and creative specs. It correlates performance metrics (CTR, CVR, CPA) with copy elements — headline structure, primary text length, tone, CTA type, pain point vs benefit framing, social proof usage. Then it generates variants that preserve the winning elements while changing specific variables. ## Practical example Your best performing Meta ad has a hook that leads with a specific number ("Cut your onboarding time from 3 weeks to 3 days"), uses a customer quote in the middle, and ends with a soft CTA. Claude generates 8 variants: 4 keep the number-led hook pattern but change the metric, 2 test a question-based hook with the same body structure, and 2 flip to benefit-first with the number as proof. Each variant changes one or two elements so you can isolate what's actually driving performance. ## What you get back - New ad copy variants grouped by what's being tested (hook, body, CTA) - Annotation showing which winning element each variant preserves - Recommended test priority based on which element has the most variance in your data - Platform-specific formatting (character limits respected for Google RSAs vs Meta primary text) - A/B test structure showing which variants to pair against each other ## When to use it - When top creatives start fatiguing and you need fresh variants fast - Creative brainstorm sessions where you need a starting point, not a blank page - When scaling to new audiences that might respond to different angles - RSA headline and description generation where you need 15 headlines that aren't repetitive ## Data access (Ryze MCP) This skill works best with live account data. Connect the free Ryze MCP once and Claude reads your Google Ads, Meta Ads, GA4 and Search Console directly: - claude.ai / Claude Desktop: Settings → Connectors → Add custom connector → `https://connector.get-ryze.ai/mcp` - Claude Code: `claude mcp add ryze --transport http https://connector.get-ryze.ai/mcp` - Cursor: Settings → MCP → add the same URL Setup guide: https://www.get-ryze.ai/how-to-connect-claude-to-google-meta-ads-mcp
[{"step":"Gather your top-performing ads. Export the last 3 months of ad data from your [PLATFORM: Google Ads or Meta Ads] account, filtering for ads with CTR > [YOUR BENCHMARK, e.g., 3.5%] and conversions > 50. Save the top 5 ads by performance.","tip":"Use the platform’s built-in 'Breakdown by Ad' report to quickly identify top performers. Focus on ads with consistent performance across multiple campaigns."},{"step":"Identify winning patterns. For each ad, extract the hook type, CTA, messaging angle, and format. Note any recurring themes (e.g., '70% of top ads use urgency-driven CTAs').","tip":"Use a spreadsheet to categorize patterns. Tools like Google Sheets or Airtable can help organize this data systematically."},{"step":"Generate variants. Paste the patterns into the prompt template, replacing [PLACEHOLDERS] with your specific data. Request 10-15 variants to ensure enough options for testing.","tip":"Specify the elements you want to prioritize for variation (e.g., 'focus on testing different hook types'). This ensures the AI doesn’t overlook key areas."},{"step":"Review and refine. Analyze the generated variants for alignment with your brand voice and audience. Adjust any variants that feel off-brand or overly generic.","tip":"Use the AI’s expected CTR and conversion potential as a starting point, but trust your intuition for brand fit. Run a small test (e.g., 3-5 variants) before scaling."},{"step":"Test and iterate. Launch the variants in your ad platform, using A/B testing to compare performance. Use the insights to refine future ad copy.","tip":"Track metrics like CTR, conversion rate, and cost per conversion. Double down on what works and iterate on what doesn’t."}]
No install command available. Check the GitHub repository for manual installation instructions.
git clone https://github.com/irinabuht12-oss/marketing-skills/tree/main/skills/ad-copy-variant-generatorCopy the install command above and run it in your terminal.
Launch Claude Code, Cursor, or your preferred AI coding agent.
Use the prompt template or examples below to test the skill.
Adapt the skill to your specific use case and workflow.
Analyze the top 5 performing ads from my [PLATFORM: Google Ads or Meta Ads] account for [PRODUCT/SERVICE NAME]. Extract the winning patterns in hooks, CTAs, messaging angles, and ad formats. Then generate 10 new ad variants that follow these patterns while introducing meaningful variations for A/B testing. Prioritize variations in [SPECIFIC ELEMENTS: e.g., hook types, emotional triggers, or value propositions]. Include metrics like expected CTR and conversion potential based on the patterns observed.
Based on your Meta Ads account for 'EcoGlow Skincare,' the top-performing ads (CTR: 4.2%-5.8%) share these winning patterns:
1. **Hooks**: 60% use 'Before/After' storytelling (e.g., 'Wrinkles in 2 weeks? Not with EcoGlow'), 30% leverage FOMO (e.g., 'Only 12 bottles left!'), and 10% use curiosity gaps (e.g., 'Your skin is lying to you').
2. **CTAs**: 80% use urgency-driven CTAs ('Shop Now,' 'Limited Stock'), while 20% use benefit-focused CTAs ('Glow in 7 Days').
3. **Messaging Angles**: 50% focus on anti-aging, 30% on natural ingredients, and 20% on affordability.
4. **Formats**: 70% are carousel ads, 20% are single-image ads, and 10% are video ads (15-30 seconds).
Here are 10 new ad variants following these patterns with meaningful variations:
**Variant 1 (Carousel - Anti-Aging Hook)**
*Image 1*: Woman with visible wrinkles before using EcoGlow.
*Caption*: 'Wrinkles in 2 weeks? Not with EcoGlow.'
*CTA*: 'Try the #1 Anti-Aging Serum Now – 30% Off Today!'
*Expected CTR*: 5.1% | *Conversion Potential*: High (aligns with top-performing angle).
**Variant 2 (Single-Image - FOMO Hook)**
*Image*: EcoGlow bottle with 'Only 12 left!' overlay.
*Caption*: 'Don’t miss out – our best-selling serum sells out fast!'
*CTA*: 'Claim Yours Before It’s Gone!'
*Expected CTR*: 4.8% | *Conversion Potential*: Medium (FOMO is a strong but saturated angle).
**Variant 3 (Video - Curiosity Gap Hook)**
*Video*: Quick zoom-in on a woman’s dull skin, then cut to her glowing skin after using EcoGlow.
*Caption*: 'Your skin is lying to you. See the truth.'
*CTA*: 'Watch the Full Transformation!'
*Expected CTR*: 4.5% | *Conversion Potential*: High (video outperforms static by 22% in your account).
... [7 more variants covering natural ingredients, affordability, and format variations].
**Recommendations**:
- Test Variants 1 and 3 first (highest expected CTR + conversion potential).
- Avoid overusing FOMO hooks (Variant 2) unless paired with a unique angle (e.g., 'Exclusive for [Your City] Only!').
- Rotate carousel images to highlight different product benefits (e.g., one image per key ingredient).skills-collection
Take a free 3-minute scan and get personalized AI skill recommendations.
Take free scan