The digital advertising landscape has undergone a seismic shift. For years, the gold standard for Meta and TikTok performance was simple: create hundreds of minor variations of a single ad concept and let the algorithm determine the winner. However, with the rollout of Meta’s "Andromeda" update, that era has effectively ended. The algorithm now treats nearly identical variations as a single creative, forcing brands to pivot from quantity of iterations to a diversity of high-quality concepts.
For small teams and lean marketing departments, this presents a daunting paradox: How do you maintain the high-volume output required by modern algorithms without succumbing to design burnout or prohibitive production costs?
Fraser Cottrell, CEO of the direct-to-consumer agency Fraggell, suggests the answer lies not in more human hours, but in a sophisticated, systematic integration of generative AI. Far from being a shortcut for the lazy, true AI-powered creative requires deep strategic input. By leveraging AI to synthesize brand identity and consumer pain points, brands can now produce professional-grade visual assets at a fraction of the traditional cost and time.
The Misconception of "Low-Quality" AI
The primary barrier to AI adoption in ad agencies is a lingering skepticism regarding quality. Many marketers still view AI-generated imagery as "cheap" or "unprofessional." Cottrell challenges this, noting that current generative models have reached a level of fidelity nearly indistinguishable from professional photography.
The perceived "low quality" of AI, he argues, is almost always a failure of the input, not the tool. AI is an amplifier; it is only as effective as the context, constraints, and instructions provided by the user. When treated as a creative partner rather than a "push-button" solution, AI levels the playing field, allowing small e-commerce brands to compete with the high-production budgets of industry giants.
Step 1: Deep Research and the Brand Knowledge Base
Before generating a single pixel, an advertiser must build a "Brand Knowledge Base." AI is a blank slate; it lacks the proprietary nuances that define a brand’s relationship with its customers.
Cottrell recommends a "Deep Research" phase using Large Language Models (LLMs) like Google Gemini. Unlike a standard search, Deep Research prompts the AI to browse the internet extensively, aggregating Reddit threads, consumer forums, and competitive landscapes. The objective is to identify the "Why": Why do customers purchase this product, and—more importantly—what specific objections prevent others from buying?
The Research Workflow
- Voice-Driven Initiation: Use tools like Whisper Flow to dictate a complex research brief into an LLM.
- External Profiling: Instruct the AI to map out the geographic concentration of the customer base, common complaints, and the core pain points that trigger a purchase decision.
- Verification: Once the AI returns a comprehensive document, it must be audited. A highly effective technique is to paste the document into Claude and instruct it to "interrogate" you. Have the AI ask questions about the data to confirm accuracy, allowing you to fill in the gaps with proprietary internal knowledge that the internet cannot provide.
This final document—a synthesis of AI-scraped public data and internal company insights—becomes the "Source of Truth" for all future AI creative tasks.
Step 2: Training a "Claude Project"
Once the research is validated, the next phase is to centralize this intelligence. Claude Projects provide a dedicated, persistent environment where the AI retains context across multiple sessions. This is the "brain" of your ad creative engine.
Essential Components of the Project Workspace:
- The Deep Research Document: The foundation built in Step 1.
- Voice-of-Customer (VoC) Data: Exported reviews and testimonials. Language used by actual customers is infinitely more persuasive than copy generated in a vacuum.
- The Internal Brand Bible: A document defining the brand’s voice, visual style, and specific definitions of what constitutes a "good" ad.
- Performance Data and Visual Context: Using tools like Poppy, marketers can analyze past high-performing ads. By feeding video files and performance metrics into the project, the AI learns which visual pacing, hooks, and on-screen actions drive conversion, turning historical data into actionable creative direction.
Step 3: Execution—The Hybrid Creative Workflow
With the Claude Project trained, the production process shifts from "creation" to "curation."
Static Image Generation
Cottrell advocates for a "Hybrid Approach" to static images. Use AI to generate the high-quality visual background or product context, but overlay the text manually. This allows marketers to swap out headlines and call-to-action (CTA) buttons without needing to regenerate the entire image.
The prompt strategy is critical:
- Ask the AI to brainstorm headlines first.
- Provide feedback on the output: "I like these two, but these two feel off-brand because…"
- Because the Claude Project maintains memory, it learns from this feedback, refining its future suggestions to better align with the brand’s specific tone.
The Video Ideation Process
While AI video generation is still maturing, its utility in scriptwriting and concepting is already transformative. Using the same Claude Project, a marketer can describe a scenario—such as a UGC (User-Generated Content) creator running a marathon—and receive a timestamped, structured script.
While a human writer is still essential to inject the final emotional nuance and conversational "spark," AI-driven scripts provide the framework, saving roughly 30% of the initial drafting time. This allows the creative team to focus on high-level strategy rather than the drudgery of drafting initial layouts.
Implications for Modern Advertisers
The shift toward AI-assisted creative is not merely a tactical update; it is a structural necessity. Meta’s Andromeda update marks a transition where the algorithm prioritizes concept diversity over iterative volume.
For agencies and in-house teams, the implications are three-fold:
- Efficiency: The time-to-market for a new ad campaign is slashed. A creative team can now test three distinct visual directions in the time it previously took to build one.
- Data-Driven Creative: By feeding actual customer complaints and performance data into the creative engine, ads become more targeted. The "guesswork" of creative direction is replaced by a system that understands exactly what the consumer wants to see.
- Scalability: Small brands no longer need to hire a full production studio to launch a high-quality campaign. They can now scale their output to meet the algorithm’s demands while maintaining a consistent brand voice.
Conclusion: The New Creative Standard
The future of ad creative is not about replacing human talent, but about augmenting it. By building a knowledge base, training a persistent AI project, and executing through a hybrid creative process, marketers can reclaim their time and significantly improve ad performance.
As Fraser Cottrell emphasizes, the most successful brands will be those that treat AI as a partner in deep research and strategic ideation. The barrier to entry has never been lower, but the requirement for strategic oversight has never been higher. By adopting this three-step system, brands can ensure they are not just producing more ads, but better ads—ones that resonate with their audience and thrive in the new era of algorithmic advertising.
About Fraser Cottrell
Fraser Cottrell is the CEO of Fraggell, a direct-to-consumer ad creative agency specializing in Meta and TikTok performance. A thought leader in the space, he shares his expertise through the Ad Creative Course and the Nice Ads newsletter, providing marketers with the tools needed to navigate the evolving digital advertising landscape. You can follow his work on X, YouTube, and through the AI Explored podcast.
