AI Stack for B2B Ad Copywriting

Winning ad performance data + ICP research → Claude → Ad copy variants in Notion + Google Ads
Matt Danese
Senior Demand Generation Manager. These stacks are built and used in production — not generated for a listicle.

B2B ad copy is usually written from intuition — someone sits down, looks at the brief, and writes four variants hoping one lands. This stack grounds the process in data. Pull your highest-performing ad copy from the past 90 days, identify the patterns — headline angles, value propositions, CTA structures — that correlate with low CPL and high MQL quality, and run the analysis through a Claude ad copy variant generator. The output is a set of structured copy variants built on what already works in your specific account — not generic B2B copy formulas from a blog post.

The Stack

Input
Winning ad performance data ICP research
AI
Claude
Output
Ad copy variants in Notion Google Ads

The Prompt

This stack is built around the Ad Copy Variant Generator Prompt. Here's the abbreviated version — the full prompt with all variables and usage notes is on its own page.

Claude Prompt — Abbreviated
You are a B2B direct response copywriter generating ad copy variants grounded in performance data.

Review the winning ad copy examples and ICP research provided below.
Identify the headline angles, value prop structures, and CTA patterns that correlate with lowest CPL and highest ICP lead quality.
Generate 5 ad copy variants for each format: headline (30 chars max), description line (90 chars), and long-form description (150 chars).
Each variant should test a distinct angle: pain-focused, outcome-focused, social proof, specificity, or urgency.
Output the variants in a structured table ready to paste into Google Ads or LinkedIn Campaign Manager.
[ ... continued — see full prompt ]

The Workflow

  1. Pull winning ad copy and performance data

    Export the top 10 performing ads from the past 90 days — by CPL and ICP-qualified lead rate — from Google Ads and LinkedIn Ads. Include the actual ad copy, not just the performance metrics.

  2. Document the ICP research

    Compile your ICP research input: target job titles, primary pain points, the specific outcome your product delivers, and any voice-of-customer language from sales call notes or G2 reviews.

  3. Identify the winning patterns

    Before running the prompt, manually review the top ads and note the common patterns: which headline angles appear in the top performers, which CTAs have the highest click-through, which value props correlate with qualified leads.

  4. Run the copy variant generator

    Paste the winning ad examples, performance data, ICP research, and pattern notes into the Ad Copy Variant Generator prompt. Claude generates 5 structured variants per format, each testing a distinct angle.

  5. Stage and test the variants

    Log all Claude-generated variants in Notion with the angle being tested and the performance hypothesis. Upload the top 3 variants per format to Google Ads or LinkedIn for live testing. Use the A/B Test Hypothesis Generator to set up the test properly.

What This Replaces

Related Stacks

New stacks drop weekly.

Each one includes the tools, the Claude prompt, and the workflow logic. Free — built for in-house B2B demand gen managers.

← Back to Stack Library

© 2026 The Demand Engineer · thedemandengineer.com