In recent years, many advertisers have mastered Meta’s targeting tools Andromeda & GEM: custom audiences, detailed interests, exclusions, and lookalike audiences. Everything felt controllable through precise settings. Today, Meta has taken over the main control of the ad delivery process.
If you continue to apply the old tactics, you risk increasingly weaker results: rising cost per result, inconsistent conversions, and the feeling that your budget is being spent without real control. This is not just a perception. The algorithm has fundamentally changed.
Meta just took the steering wheel out of your hands.
In the last two years, Meta has introduced two new artificial intelligence engines behind its advertising system — developments that represent an advanced stage in the evolution of Programmatic Advertising toward greater automation and personalization.
Andromeda is the first engine — an advanced ad retrieval system. When a user opens Instagram or Facebook, Andromeda analyzes tens of millions of possible ads in milliseconds and creates a highly relevant shortlist for that specific user, even before the auction stage begins.
GEM (Generative Ads Recommendation Model) is the second engine and the main driver of change in 2026. Trained on billions of purchasing behaviors, GEM accurately predicts who is likely to convert, not just who interacts superficially. It functions as an extremely sophisticated recommendation system that surpasses the capabilities of manual targeting.
Since the wide-scale implementation of GEM, Meta has reported conversion increases of approximately +5% on Instagram and +3% on Facebook Feed. At the platform’s volume, these percentages represent significant impact and confirm the superiority of the automated system over traditional manual optimizations.
This is why broad audiences now frequently outperform manually crafted ones, and why last year’s settings have become ineffective. You didn’t misconfigure anything — the platform has evolved.
The key strategic question is no longer “How do I target?”, but “What am I feeding the system?”The answer involves two essential elements:
1- Clean Data Signals
Conversion events must be set up correctly so the model can accurately identify your ideal customer profile. Inaccurate data leads to flawed optimization and budget waste.
2- CREATIVITY– The New Decisive Targeting Factor
Ads (images, videos, and text) are now read and interpreted by AI. Through Andromeda and GEM, the platform matches the creative with the most relevant users. This is why creative quality and diversity have become more important than targeting settings.
While the general advice to test multiple variations remains valid, it is not the raw volume that matters, but real diversity. Meta’s algorithm automatically detects and groups creatives that are visually or conceptually similar, treating them as essentially the same ad.
Real diversity means creating ads that communicate different messages, solve different problems, or speak to different stages of the customer journey.
For example, if you run a restaurant and post daily photos of plated dishes, Meta will see them as very similar creatives, regardless of how many you upload.
The same happens when you generate dozens of images in bulk using tools like Canva — even with different filters or slight layout changes. These are still considered minor variations of the same concept. As a result, Meta will only deliver a few of them, ignoring the rest.
So if you have 10 photos with food – even there are different angles, META will dont think there are 10 phtos but only one!!
For optimal results, each new creative should differ significantly in at least one of the following aspects:
- The problem it solves (e.g., stress relief vs. time saving vs. money saving)
- The audience’s stage of awareness
- The delivery format (person-led video, clean static image, testimonial, etc.)
- The initial hook (the first element that stops the scroll)
The goal is NOT to produce dozens of ads, but to identify the one that drives the majority of profitable results. A single exceptional creative can support consistent daily spending while remaining profitable, outperforming all other variations combined.
Producing variations is a technical operation. Creating “Kreatives” that turn an ad into a long-term profitable asset is a true strategic craft — backed by research, relevant angles, strong hooks, and a clear offer.
In the current context of Programmatic Advertising, success belongs to those advertisers who understand that although direct control over targeting has diminished, the quality of signals and creatives remains firmly in their hands.
Have you already noticed these changes in your Meta campaigns in 2026? Feel free to share your experience in the comments below.