Chris Long on Honest Ecommerce mentioned using ChatGPT as a baseline for every marketer’s output, and this is the perfect mental model for creative testing today. It means you treat AI not as a magic button for finished ads, but as an engine for generating a wide pool of creative ideas and assets you can then go test.
The first step is mass ideation. Instead of a few marketers brainstorming in a room, you can use AI to generate hundreds of hooks, angles, and copy variations in minutes. What problems does your product solve? What are the top three emotional triggers for your customers? You can feed these prompts to a tool and get a huge amount of raw material. This isn't about finding a perfect line of copy, but about exploring the entire landscape of possibilities so you don't miss a winning angle. This is a core theme in the broader conversation around AI in e-commerce, where AI helps expand what's possible.
Next, you move to asset generation. This is where you can turn those ideas into actual creative. On an episode of New Frontier, Leo Sgovio talked about how to use AI-generated UGC for sustainable growth. This is a powerful tactic. You can use AI to create visuals that have the look and feel of user-generated content, which is often more trusted and effective on platforms like Meta. This allows a brand to produce a massive volume of authentic-looking content without needing to coordinate with hundreds of creators. It drastically lowers the cost and time required to produce creative for testing.
Of course, AI output requires a human filter. Chris Long also made the crucial point on his episode that you can’t just take what an AI gives you at face value. A human marketer needs to refine the ideas, check the copy for brand voice, and ensure the visuals are high-quality and accurate. The AI is a powerful assistant, but you are still the creative director. Your job is to curate, edit, and select the best of what the AI produces, ensuring it aligns with brand strategy before it ever gets a single dollar of ad spend.
Where this framework breaks down is in the actual execution of the test on Meta's platform. The hosts I listened to are focused on the inputs of the creative process, the ideation and generation of assets. They don't go into the tactical details of whether to use an Advantage+ campaign, a dynamic creative test, or a manual ad set structure with isolated variables. This AI-first production model gives you the fuel, but it doesn't tell you how to build the engine.