What Are Meta AI Systems? GEM, Lattice, Andromeda, and More

Meta’s latest set of AI systems finished rolling out at the end of 2025, and totally changes how paid social operates in the process. Here’s why creative, messaging, hooks, angles, offers, and diversification allow the platform to keep connecting your campaigns with the right people.

It’s not a stretch to say that, over the past few years, artificial intelligence has shifted the foundation of digital marketing in some major ways.

Nowhere is that shift more evident than at Meta. From algorithmic automation in ads to creative optimization and generative tools, Meta’s AI systems are drastically changing how campaigns behave, audiences are reached, and performance is measured.

The latest evolution — Meta Andromeda — isn’t just another feature rollout. It represents a structural shift in how Meta leverages AI across its advertising platform.

With all of these new shifts, changes, and systems launching, we wanted to help digital marketers make sense of it so they can better prepare and move forward with confidence. For help, we chatted with HawkSEM Senior Lead Strategist Nicole Goodnough.

(Image: Unsplash)

Meta GEM

Meta GEM acts like a super brain, quickly analyzing and connecting trillions of data points across all types of activities.

It catalogs and understands complex relationships between information, enabling Meta’s recommendation system to learn from massive datasets and recognize subtle patterns.

This allows Meta to provide the most relevant ads to the right person at the right time, with minimal delays.

GEM makes Meta’s ad system incredibly intelligent, ensuring that ads are highly targeted and timely based on real-time data analysis.

Meta Lattice

Think of Lattice like a giant library.

Meta Lattice is a unified system that integrates information across multiple domains, replacing the need for several smaller, subject-specific systems.

In the past, each ad product was handled by a separate library of information, but Meta Lattice combines these into one powerful system, learning from all available data.

This streamlined approach improves ad relevance and effectiveness by applying cross-domain knowledge, ensuring faster and more accurate ad delivery.

Instead of multiple fragmented systems, Meta Lattice enhances efficiency and ad targeting by drawing from a single, comprehensive source of data.

Meta Andromeda

Basically, Andromeda is the personal concierge AI system.

Imagine having an assistant who knows your tastes so well that they don’t just understand that you covet shoes, but that you like to wear red flip flops at the beach.

Meta Andromeda learns your preferences, so Meta can show you ads that are more relevant and interesting.

Meta Sequence Learning

Sequence Learning is like a memory game.

With traditional aggregated data models, a user would continue to see other ski resort ads if they converted on one ski resort ad, for example.

With recent changes in Meta’s ads learning model, after purchasing a ski resort room, a person would now see ads for ski equipment, lift tickets, or ski luggage, providing more relevant ads personalized to the purchase journey.

What Meta’s ad updates mean for digital marketers

“I don’t think marketers really understood the impact these updates would have on Meta ads,” says Goodnough.

“GEM, for example, is Meta’s ‘super brain’ — that means it analyzes trillions of bits of information and is able to make personalized recommendations based on subtle patterns in online behavior so users see ads that are more relevant to them.”

As a result, she says these systems have drastically changed the entire way paid social operates, essentially causing audience targeting to be moot.

“As advertisers, we tend to want to maintain a sense of tight control over things so we know exactly what levers to pull to drive success for our clients,” she explains.

“Meta’s AI systems take away a lot of those levers that we used to need control over, pushing us to learn more ways to gently nudge the platform in the right direction instead.”

On the plus side, digital marketers are still able to accomplish those same tasks through creative: using different messaging, diversifying content, and speaking to the people on the receiving ends of ads in a more direct manner.

While campaign settings and audiences are broader, Goodnough says creative should be more specific. That gives Meta the ability to analyze as much data as possible.

“We [can say to Meta], ‘now that you have a ton of data, this is what I’m selling, so go find the best people in this audience,’” she explains.

“Not only did these changes alter the way we run large prospecting campaigns, but they also fundamentally changed how we need to think about the funnel.”

(Image: Unsplash)

5 ways digital marketers can move forward

All this info got your head spinning? Don’t throw out your entire paid social strategy just yet.

Instead, use this info and the expert tips below to refine and modify your plan for best results in the months to come.

1. Reframe your campaign strategy

• It’s wise to build campaigns with broader audiences (such as interest cohorts, lookalikes, and value segments) rather than hyper-specific lists.
• Use higher-level objectives (like Purchases and Value) instead of intermediaries like Link Clicks.
• Allocate budget based on performance ecosystems, not individual ad sets.

2. Invest in your creative

• Produce diverse creative banks with multiple hooks, lengths, and formats.
• Leverage dynamic creative optimization (DCO) tools if applicable.
• Pair human insights with AI to refine your messaging — machine learning can optimize creative, but it needs meaningful variations to learn.

3. Prioritize data and signal quality

• Remember that AI performance degrades without reliable signals.
• Ensure Conversions API is configured and firing events properly.
• Use enhanced conversions where possible.
• Audit your conversion events and ensure they’re mapped correctly to business value.

4. Rethink your KPIs

• Set KPIs that reflect value, not vanity (like revenue per impression rather than CTR).
• Use predictive metrics like Cost per Estimated Action (CPEA) to understand directionality before conversion windows close.
• Look at signal velocity — the speed at which your campaigns generate meaningful data.

5. Conduct strategic testing

• Use controlled tests to compare different bidding strategies, audiences, and creative approaches.
• Run budget-split tests rather than audience or placement splits — let Andromeda’s AI optimize within each test cell.
• Compare long enough to capture learning phases.
• A/B testing CTA buttons or different background colors isn’t something that will move the needle in a meaningful way. Rather, test out bigger elements like different hooks, like “476 hours saved annually” if you’re a security software company, for example.
• Include different unique value propositions (UVPs) in each ad and present them differently: A mom in a kitchen making cookies who says, “This all-natural sugar substitute is perfect for sweet-tooth cravings without the calories,” against a static ad with text overlaid in front of the product saying, “Just one ingredient for any recipe — a sugar alternative for everyone.”
• Test free shipping vs. 10% off with shipping

The takeaway

Everyone’s purchase journey is unique and by no means linear. Consumers don’t shop in a funnel, and Meta helps fuel purchase decisions without us creating a top-of-funnel to bottom-of-funnel audience structure.

“Historically, if you didn’t exclude engaged users or current customers from sales campaigns, Meta would overspend on these audiences because they’re more ready to buy than cold audiences are and have a better return at first,” says Goodnough.

“This hasn’t been happening for a while, and anyone who tells you it does is stuck in 2023. While it’s difficult to relinquish the control of having a beautiful TOF > MOF > BOF structure, if you want to succeed, you have to.”

Make this the year you blow past your digital marketing goals —chat with us today to find out how we can help.