Meta shares insight into the evolving AI-powered ad targeting process

Meta has shared some new insights in the evolving ad targeting systems, and how growing AI processing power is delivering better results for advertisers through improved interest matching.

And advertisers have taken notice. More and more Meta advertising partners are reporting improved performance, with AI targeting helping customers find things they would otherwise have missed.

In his new overviewMeta provides greater insight into how the system works and how it drives broader performance improvements across Meta’s advertising offerings through continuous improvement.

As explained by Meta:

“The Generative Ads Recommendation Model (GEM) is Meta’s most advanced ad base model, built on an LLM-inspired paradigm and trained on thousands of GPUs. It is the industry’s largest basic recommendation system (RecSys) model, trained at the scale of large language models.”

To be clear, Meta has been using advanced machine targeting for ads for years, with a vast wealth of audience interest and engagement data that allows Meta to more accurately identify users’ interests and serve relevant ads accordingly.

Before the latest wave of AI tools hit the market, Meta had been using the same LLM-based approach to targeting for years, but reframing scaled data processing as “AI” has changed the paradigm around how this is perceived.

Essentially, Meta was criticized for facilitating psychographic targeting, based on the data it has on its 3 billion users, including the pages they like, people they are connected to, interests, traits, etc.

But now, not only is all this an acceptable practice, under the banner of ‘AI’, but Meta’s data is also considered a great advantage. And with this in mind, after enduring all those setbacks, you can see why Zuckerberg is so eager to claim the title as the leader in the AI ​​space.

Meta says its latest GEM model offers significant advancements in its targeting systems, by using “model scaling with advanced architecture, post-training techniques for knowledge transfer, and enhanced training infrastructure to support scalability.”

“These innovations efficiently increase ad performance, enable effective knowledge sharing across the ad model fleet, and optimize the use of thousands of GPUs for training. GEM has created a paradigm shift in advertising RecSys, transforming ad performance across the funnel – awareness, engagement and conversion – through joint optimization of both user and advertiser objectives.”

In summary: more people click on ads, more ad customers sell things.

In terms of performance specs, Meta says the updated system now:

• 4x more efficient at driving better ad performance for a given amount of data and computing power than the original ad recommendation ranking models.
• 2x more effective in knowledge transfer, optimizing broader advertising performance.
• Faster and more effective based on greater computing capacity, enabling more effective scalability of advertising results.

“GEM is trained on ad content and user engagement data from both ads and organic interactions. From this data, we infer features that we categorize into two groups: sequence attributes (such as activity history) and non-sequence attributes (such as user and ad attributes – for example age, location, ad size, and creative representation). Custom attention mechanisms are applied independently to each group, while also enabling cross-feature learning. This design improves accuracy and scales both the depth and breadth of each attention block, allowing 4x the efficiency of our previous generation models.”

So Meta’s advertising system now has a more systematic capacity, allowing it to process more information and find more correlating data characters, leading to improved advertising performance.

This is also reflected in the performance data.

Meta previously shared that advertisers using its various AI-powered ad targeting options have seen significantly improved ad performance, while it also revealed plans to eventually automate the entire ad creation processusing these evolving systems to create your ad, optimize your targeting, and manage your budget without having to do anything other than enter your product URL.

That’s how much confidence Meta has in its advertising systems to drive better performance over time.

Meta’s GEM system works together with Meta’s “Rooster’architecture, and being’Andromeda” models, each of which plays its own role in optimizing your meta ad targeting.

• Lattice is what Meta calls its “ad library,” which determines ad ranking and ensures optimal placement for each campaign
• Andromeda is Meta’s personalization model, which guarantees the relevance of ads based on each user’s engagement history and interests

Combined, these systems deliver greater ad relevance, using Meta’s ever-expanding suite of technologies to learn more about each user’s preferences and improve targeting accordingly.

Which, again, at Meta’s scale means a lot of data points to process, which can lead to highly accurate, highly valuable advertising results.

I mean, in 2015, reports suggested that Facebook already had enough data to infer pretty much everything about you, based on your in-app activity.

That capacity is boosted by the latest AI models, leading to better ad performance across the board.

It’s interesting to consider Meta’s capabilities in this regard, and it might be worth trying out Meta’s evolving AI-powered advertising options, via Benefit+to see what results you get.