By Emily Hopkins, Digital Marketing Manager
Meta has stepped into a whole new level of AI-powered advertising. With tools like GEM, Meta Lattice, Sequence Learning and the upgraded Andromeda system, the platform now processes huge amounts of data, predicts what people are likely to do next and selects the most relevant creative in seconds.
GEM quickly identifies which creatives are likely to perform best.
Meta Lattice learns across all campaign objectives at once.
Sequence Learning tracks what people do before and after seeing an ad to predict their next step.
And Andromeda pulls it all together — narrowing millions of options down to the most relevant few thousand in real time.
Together, these updates fundamentally change how ads are delivered and what advertisers need to prioritise.
Why this matters for everyday advertisers
These changes impact day-to-day performance. The old approach of relying heavily on manual targeting and rigid control just isn’t as effective anymore. Meta’s AI is now far more influenced by behaviour, creative quality and quantity and overall user intent.
If your ads lack variation, feel too promotional or don’t offer enough value, the system is more likely to deprioritise delivery. Understanding how AI works helps you manage performance and avoid those frustrating dips that seem to come out of nowhere.
Creative is now the new targeting
Here’s the shift in simple terms:
Your creative now does most of the targeting for you.
Meta’s AI works best when advertisers:
- Mix different styles, hooks and messages
- Refresh creatives frequently
- Use a blend of formats — videos, stills, carousels, UGC-style content
- Create content that feels human, helpful and engaging
Every new creative gives the system a better chance of matching the right message with the right person.
Why you need much more creative variation
Not long ago, the advice was “don’t overload campaigns with too many ads” because it slowed learning. That’s no longer the case.
With Andromeda and GEM, more creative variation actually improves performance. Meta can now handle huge creative libraries without slowing down — and the more options you give it, the better it performs.
This shift is especially important for CBO (Campaign Budget Optimisation) campaigns. CBO used to reward simplicity and fewer ads per ad set. But now, Meta expects CBO campaigns to include a far broader mix of creative styles, formats and messages. A single CBO campaign should now feel more like a creative library — not a small, tightly controlled set of ads.
In short:
CBO + variety = better learning, faster stability and stronger overall results.
What Meta expects from advertisers now
To work effectively with Meta’s AI, advertisers should:
- Use broad targeting and Advantage+ structures where possible
- Prioritise high-quality signals (good conversion data, consistent events)
- Build campaigns that support each other across objectives
- Adopt a creative-first mindset instead of a targeting-first approach
- Refresh and test creative regularly
- Provide significantly more variations so the system has enough choice to optimise properly
Brands that thrive in this new landscape are the ones producing relatable, engaging, varied creative and giving Meta’s AI plenty to work with — not those relying on strict targeting or a small set of evergreen assets.