- Nov 29, 2022
- 2,491
- 285
Meta first introduced Andromeda, marking the initial shift away from the familiar model of granular campaign control. Then came GEM — a new engine that fundamentally changed the rules of the game. It increasingly looks like the era of precisely targeting narrowly defined audiences is coming to an end.
For years, Meta has been guiding advertisers toward a model where audience selection, ad ranking, and delivery are handled almost entirely by AI-driven systems behind the scenes. In this article, we’ll examine the roles of Andromeda and GEM, how these changes affect campaign structure, and how affiliate marketers can adapt to remain competitive.
Andromeda and GEM: Who Does What
Some affiliate marketers assume that Andromeda and GEM are simply two names for the same update. In reality, they serve different functions, and understanding this distinction is critical for adapting to current market conditions.Andromeda
Andromeda is responsible for discovery. Its role is to identify, from billions of ads, those most likely to appeal to a given audience. Conceptually, it selects relevant options and places them into the system for further evaluation.GEM
GEM focuses on creative delivery. It is a full-scale large language model that analyzes user behavior over extended periods and determines how best to present an offer. For example, it may first show a short-form video, then a testimonial, and only later introduce the offer organically within the feed.In simple terms, Andromeda functions as the system’s hands, while GEM acts as its brain. The platform generates and sequences content independently, deciding how to present a product based on user behavior. As a result, the primary focus shifts from technical campaign settings to creative strategy. The key requirement is supplying creatives that the AI can combine into a coherent journey aligned with user preferences. According to Meta’s developers, this approach has already increased average conversion rates by approximately 5% on Instagram and 3% on Facebook.
How the New System Impacts Campaign Setup
With the introduction of GEM, traditional approaches to campaign preparation and management are changing. To clarify how to operate under the new model, it’s important to look at the core shifts in more detail.Facebook and Instagram Are Now Fully Integrated
Before GEM, Meta maintained separate user databases for Facebook and Instagram. Even if the same individual actively used both platforms, the system treated them as two distinct users.Now, internal systems build a unified behavioral and preference profile, allowing the platform to determine where an offer is most likely to perform best. If a user consistently engages with short-form content related to fitness, dieting, and weight loss, GEM may surface user-generated content in their feed related to relevant wellness products.
Over time, manual placement selection is likely to become obsolete. Restricting placements can interfere with the system’s ability to identify the most effective delivery path for an offer.
Ads Must Feel Native
GEM has moved beyond simple keyword recognition and can now interpret overall user intent. The model understands preferred content types and formats, such as memes, user-generated content, or short videos. As a result, Andromeda is more likely to prioritize campaigns from affiliate marketers who understand their audience’s content preferences.Previously, heavily stylized or overtly promotional creatives could still perform. Today, this approach is far less effective. Creatives must closely resemble the content users already consume. Otherwise, even high-quality production will struggle to deliver results.
Traditional Creative Testing Is Becoming Less Effective
To identify a converting setup, many affiliate marketers historically tested variations within a single concept — changing calls to action, buttons, or visual details. GEM, however, evaluates creatives at a semantic level. From its perspective, dozens of celebrity-based creatives may appear effectively identical.To gain an advantage, marketers need to test fundamentally different creative angles: authority, benefits, emotional appeal, social proof, or user-generated formats. The broader the range of approaches provided to GEM, the more effectively the system can match offers to different audience segments.
How to Adapt to the New Reality
Strategies that worked in the past are becoming increasingly ineffective. Meta continues to centralize control over campaign mechanics to lower entry barriers, expand advertiser adoption, and reduce policy violations. As AI investment increases, manual control becomes more limited. Experienced affiliate marketers recommend the following adjustments.Simplify campaign structure
Overly narrow targeting limits algorithmic flexibility. Creating numerous ad sets with minor creative differences often does more harm than good. A more reliable approach is to focus on high-quality creatives built around clearly distinct concepts.Reduce manual intervention
Frequent budget changes, restarts, and extensive automation rules disrupt GEM’s learning process. Meta allows its models to analyze user behavior over several months. Abrupt changes reset accumulated insights and force the system to relearn from scratch.Plan for longer learning cycles
When relying on automated campaign types, meaningful performance evaluation within the first 24 hours is unrealistic. The system typically requires three to four days to determine optimal audience distribution. This necessitates greater patience and a more flexible budget strategy.Duplicate-based scaling is becoming less effective. Instead of replicating campaigns, it is often more productive to enhance a single performing ad set by introducing new creative formats and angles. Ideally, GEM should receive not just a collection of creatives, but a structured sequence of assets that can be combined and adapted for different audience segments.