Scaling Ad Creative in the Age of AI: A Strategic Framework for Modern Marketers

The digital advertising landscape is currently undergoing a seismic shift, driven by the dual pressures of algorithmic evolution and the rising cost of traditional content production. As Meta’s ad delivery systems become increasingly sophisticated, the demand for high-volume, high-quality, and—most importantly—genuinely distinct creative assets has reached a breaking point for many small-to-medium-sized brands.

Fraser Cottrell, CEO of the direct-to-consumer (DTC) ad creative agency Fraggell, argues that the traditional model of ad production is no longer sustainable for brands looking to compete. In a recent deep dive into the intersection of generative AI and performance marketing, Cottrell, alongside Michael Stelzner, outlined a comprehensive three-step system designed to harness artificial intelligence not as a shortcut for the "lazy," but as a sophisticated engine for creative scale.

Main Facts: The New Reality of Digital Advertising

The primary catalyst for the adoption of AI in ad creative is the recent "Andromeda" update to Meta’s algorithm. Historically, advertisers could find success by running hundreds of slight iterations of the same creative—changing a background color or a single word in a headline. However, Meta now groups these minor variations together, treating them as a single creative entity. To achieve "breakout" success, brands must now provide the algorithm with fundamentally different creative concepts.

This shift presents a significant barrier to entry for smaller brands that lack the budget for continuous studio shoots or large-scale design teams. Cottrell identifies two major misconceptions that prevent marketers from leveraging AI to solve this problem:

1. The "Laziness" Myth: Many believe using AI is a way to bypass the hard work of marketing. On the contrary, Cottrell asserts that high-quality AI output requires significant effort in research, context-setting, and prompt engineering.
2. The Quality Barrier: While early AI models produced uncanny or low-quality images, modern models like Gemini and Nano Banana produce static imagery that is virtually indistinguishable from professional photography. While video generation is still maturing, it has already become an indispensable tool for ideation and scripting.

By integrating AI, the "playing field is leveled." A product shot that previously cost thousands of dollars in studio fees can now be generated for cents, allowing brands to reinvest those savings into testing a wider variety of psychological hooks and visual styles.

Chronology: The Three-Step Workflow for AI Ad Production

Cottrell’s methodology moves from deep foundational research to technical training and, finally, to the execution of creative assets. This systematic approach ensures that the AI isn’t just guessing but is operating as an informed extension of the brand’s marketing department.

Phase 1: Building the Brand Knowledge Base through Deep Research

The process begins not with a prompt for an image, but with an exhaustive investigation into the brand’s ecosystem. Using the "Deep Research" functions available in Large Language Models (LLMs) like Google Gemini, Cottrell constructs a comprehensive external profile of the brand.

The objective is to move beyond surface-level marketing speak. The AI is instructed to browse the internet—specifically forums like Reddit and review sites—to identify:

• The "Why": Why are customers buying the product?
• The "Why Not": Why did potential customers choose a competitor or abandon their purchase?
• Pain Points: What are the recurring complaints or objections?
• Geography: Where is the customer base concentrated?

To ensure the prompt for this research is airtight, Cottrell often uses voice dictation (via tools like Whisper Flow) to describe his needs to Claude, which then generates the technical prompt for Gemini. Once the research is generated, it is "vetted" by Claude to identify hallucinations and corrected with proprietary internal data that the internet cannot access, such as customer support insights or internal sales data.

Phase 2: Training the "Brain" (The Claude Project)

The second phase involves moving the verified research into a "Claude Project"—a dedicated workspace with persistent memory. This acts as the "Source of Truth" for the brand.

Cottrell populates this workspace with:

• The Deep Research Document: The refined output from Phase 1.
• Voice of Customer (VoC) Data: Exported reviews and testimonials.
• Internal Brand Guidelines: Documents outlining the brand’s mission and what constitutes a "good ad" for their specific niche.
• Performance Data: Historical data from Meta Ad Manager.

A critical component of this training is the use of visual analysis. Cottrell uses tools like Poppy or Gemini’s native vision capabilities to "watch" top-performing videos. The AI then summarizes the pacing, hooks, and visual elements that led to success, adding these qualitative insights to the project’s memory.

Phase 3: Creative Execution and Iteration

With the AI fully briefed, the creative process begins. Cottrell advocates for a "Hybrid Approach" to static imagery: using AI for the visual and headlines, but performing the final layout and text overlay manually.

1. Copywriting: Claude generates headlines based on the project data. The user provides feedback (e.g., "I like these two, I hate these two because they are too aggressive"), which Claude uses to refine its style over time.
2. Image Prompting: Claude is asked to brainstorm visual concepts that align with the chosen headlines. Once a concept is selected, Claude writes a technical prompt for an image generator.
3. Generation: The prompt is taken to a tool like Nano Banana 2 Pro (within Gemini). Cottrell recommends generating images on clean, brand-aligned backgrounds and maintaining a "real-world" aesthetic to avoid the "plastic" look often associated with AI.
4. Video Scripting: While full AI video production is not yet agency-standard for Fraggell, AI is used to generate 30-second UGC (User Generated Content) scripts. These scripts serve as a "30% head start," providing a framework that human writers can then refine for nuance and emotional resonance.

Supporting Data: The Economics of AI-Driven Creative

The shift toward AI is driven by a clear economic incentive. In the traditional agency model, a single product photoshoot might yield 10–15 usable assets. If those assets fail to resonate with the Meta algorithm, the brand has "burned" its budget with no immediate recourse.

In contrast, the AI workflow allows for:

• Cost Reduction: Generating a high-fidelity product image via Gemini or Midjourney costs a fraction of a cent in API fees.
• Speed to Market: A concept can move from a Reddit-sourced pain point to a live ad in under an hour.
• Volume Requirements: To satisfy the current Meta environment, brands often need to test 5–10 fundamentally different "hooks" per week. AI makes this volume achievable for a one-person marketing team.

Furthermore, tools like Poppy and Gemini’s vision capabilities provide data-driven creative insights that were previously anecdotal. By "quantifying" the visual elements of a successful ad (e.g., "The hook occurred at 1.5 seconds with a high-contrast background"), brands can replicate success with higher mathematical certainty.

Official Responses and Expert Perspectives

Fraser Cottrell emphasizes that the role of the marketer is shifting from "creator" to "curator and director." He notes that the most successful users of AI are those who treat the LLM like a highly talented but literal-minded intern.

"AI is only as good as the context and instructions you give it," Cottrell states. He argues that the human element remains the most critical part of the chain. While AI can synthesize thousands of Reddit comments into a script, it cannot yet replicate the specific "human spark" or the cultural zeitgeist that makes a video go viral.

Industry analysts suggest that Meta’s Andromeda update was designed to force this exact behavior: a move away from "spammy" iterations and toward high-value creative diversity. By rewarding "meaningfully different" assets, Meta is essentially training its advertisers to become better storytellers—and AI is the tool that allows them to do so at the required scale.

Implications: The Future of the Creative Department

The widespread adoption of this three-step system has profound implications for the future of marketing roles.

1. The Death of the Generalist Designer? Purely execution-based design roles (e.g., "remove this background") are being replaced by "Creative Technologists" who understand how to bridge the gap between LLMs and image generators.
2. The Rise of Deep Research: As AI makes the creation of assets easier, the competitive advantage shifts back to research. Brands that understand their customer’s psychology better will write better prompts, leading to better AI outputs.
3. Hyper-Personalization: We are moving toward a world where ads are not just targeted based on demographics, but generated in real-time based on a user’s specific objections. If the AI knows a user is worried about "shipping times," it can theoretically serve a creative asset generated specifically to address that pain point.

Ultimately, the integration of AI into ad creative is not about replacing humans, but about removing the mechanical "bottlenecks" of production. By automating the research, training, and initial drafting phases, creative teams are freed to focus on high-level strategy and the emotional storytelling that truly drives consumer behavior. For the small brand, AI is no longer an optional luxury; it is the necessary infrastructure for survival in a platform-driven economy.