Meta's Generative Ads Model Is Delivering 18.9% More Conversions: What

On October 13, 2026, Meta published results from its next-generation Generative Ads Model (GEM), its largest AI recommendation system ever built. For ecommerce brands running Facebook and Instagram ads, this means an automatic 18.9% conversion improvement and 12% click-through rate improvement across Meta's ad surfaces, with gains already live in every account today.

What Does Meta's Generative Ads Model Mean for Your Campaigns?

GEM shifts Meta's ad system from scoring every individual ad candidate separately to using a single LLM-scale foundation model that reasons about ad content and user preferences simultaneously. Campaigns running Advantage+ structures and broad targeting capture the full benefit automatically. Manual campaigns with heavily restricted targeting and placement exclusions may see smaller gains because they give GEM fewer variables to optimize.

What Is Meta's Generative Ads Model and How Does It Work?

GEM builds on Meta's Generative Recommenders (GenRecs) architecture, first introduced in 2025. The 2026 generation replaces separate per-domain optimization models, one for Facebook, one for Instagram, one for Reels, with a single central foundation model that learns from data and signals across all of Meta's ad platforms at once.

Meta calls this architecture Kunlun, co-designed with the latest GPU hardware. It enables models 2.5x more complex than the prior generation, using LLM-scale computation and long-context learning over user engagement history. The result mirrors how frontier LLMs improve with scale: more capacity and longer context produce qualitatively better rankings that translate into real revenue outcomes for advertisers.

On October 19, 2026, Meta published a companion paper on GenRecs Tokenization, the foundational data representation that powers GEM. Rather than maintaining a dedicated ID for every ad, GEM now represents each ad using a stable semantic vocabulary of tokens. This makes Meta's system significantly better at surfacing new products with no prior engagement history, which is critical for ecommerce brands launching new SKUs or testing new audiences ahead of Black Friday.

The results across both papers are significant: 18.9% conversion improvement and 12% CTR improvement overall, with surface-specific lifts of 5.8% on Facebook, 4.1% on Instagram, and 1.6% for off-site conversion events. GenRecs Tokenization added an additional 0.35% conversion lift across surfaces. These gains are fully automatic and require no opt-in or configuration changes from advertisers.

Which Campaign Types Are Seeing the Biggest Gains from GEM?

GEM delivers simultaneous performance improvements across every surface and funnel stage, including awareness, consideration, and deep-funnel conversion objectives. But the magnitude of the gain depends on how much signal you give the model. Campaigns that provide GEM the most input data benefit most: broad targeting, Advantage+ audience settings, Advantage+ Creative enabled, and multiple ad formats in a single ad set all increase the signal pool. Manual campaigns with tight audience restrictions, limited creative variations, and placement exclusions reduce the pool and limit GEM's optimization range.

For 7 and 8-figure ecommerce brands, this is the most concrete data Meta has ever published to justify moving toward Advantage+ structures and trusting AI-driven delivery. The performance gap between AI-first and manual-first campaigns will widen as GEM continues to improve heading into BFCM.

How Does GEM Change Your Meta Ads Strategy Before Black Friday?

GEM changes the underlying calculus for how brands should structure their accounts heading into BFCM. If Meta's AI is now significantly more accurate at matching ads to buyers, tight audience targeting provides less additional lift than it did a year ago. Creative quality and product data quality become the primary levers brands control directly. Your job is to give GEM the best possible inputs: rich, varied creative and a clean, complete product catalog. GEM handles the matching from there.

This also has implications for how you interpret performance changes in October. If your conversion rates have improved in the past two weeks without any campaign changes, GEM is the most likely driver. Avoid over-interpreting that lift as a reason to restructure campaigns. Let the model continue optimizing with the data it has.

For a complete picture of what has changed on Meta this year, see our full Meta Ads 2026 update tracker. For brands managing creative at scale, our analysis of Meta's Creative Diversity Score explains exactly what types of creative variation GEM rewards most. Our Meta Ads agency team is available for brands that want help restructuring for BFCM.

STAY CURRENT

Ad platforms change weekly. Your inbox shouldn't miss it.

Get CTC's weekly breakdown of the platform changes that actually matter for ecommerce brands -- no noise, just what affects your campaigns.

What Should You Do Now with Meta's Generative Ads Model?

Audit your Advantage+ Creative settings today. Go to Ads Manager, open your active ad sets, and confirm Advantage+ Creative is enabled. Every additional creative enhancement you allow gives GEM more variables to optimize. Brands that restrict all enhancements limit the model's ability to find the best-performing version of each ad.
Add at least 5 creative variations to each active ad set before November 1. GEM learns from creative signals across all surfaces. More variations mean more data for the model to identify which combination of image, copy, and format performs best. Multiple creative formats in a single ad set, including static, video, and carousel, give GEM the broadest optimization range.
Clean your product catalog before October 28. GEM uses product data to reason about ad content and match ads to high-intent buyers. Catalog entries with missing descriptions, outdated prices, or incomplete attributes reduce matching quality. Run a catalog diagnostic in Commerce Manager and resolve all disapproved products before product browsing auto-enables on October 28.
Run a test Advantage+ Shopping Campaign alongside your manual campaigns in October. If your current campaigns rely on heavily restricted targeting, you need real BFCM-season data from Advantage+ structures before peak. Launch a test now, not in November, so the model has time to learn.
Do not restructure campaigns in response to recent unexplained performance improvements. If conversions have increased in the past two weeks without campaign changes, GEM is the most likely driver. Wait for at least three weeks of consistent data before drawing structural conclusions or making major changes.

Frequently Asked Questions

Does Meta's Generative Ads Model change how I should set up campaigns?

GEM is a backend AI model that improves Meta's ad delivery automatically for all advertisers. You do not need to opt in or change campaign types to benefit. However, campaigns with Advantage+ features enabled and more creative variety give GEM more signal to work with, which is consistently associated with stronger performance gains.

Will GEM affect my cost per acquisition before Black Friday?

Meta's published results show 18.9% conversion improvement across surfaces. If that lift holds in your account at the same spend level, your effective CPA will decrease. However, BFCM auction competition increases CPMs significantly starting in late October, so the net effect on CPA depends on your category, competition level, and audience size.

What is the difference between GEM and Advantage+ Shopping Campaigns?

GEM is the underlying AI model that powers Meta's ad delivery across all campaign types, including manual campaigns. Advantage+ Shopping Campaigns are a specific structure that gives GEM maximum latitude in targeting, creative, and placement. Running Advantage+ does not activate GEM, which runs for all advertisers automatically, but it does maximize the input signal GEM has to optimize against.

Does GenRecs Tokenization improve how new product ads perform?

Yes. GenRecs Tokenization specifically addresses the cold-start problem for new products. Previously, new SKUs with no engagement history received lower delivery quality because the system had limited signal to work from. Semantic tokenization lets GEM reason about new products based on what they actually are, not just their historical performance. This is particularly relevant for brands launching new SKUs for BFCM.

Meta's Generative Ads Model Is Delivering 18.9% More Conversions: What Every Ecommerce Brand Must Do Before Q4