Meta is scaling its Generative Ads Model (GEM), a recommendation model designed to improve how ads are ranked and delivered across its advertising ecosystem. In an engineering update published October 13, the company described a shift toward a more centralised foundation model rather than isolated optimisations for individual surfaces such as Facebook, Instagram and Reels.

What Meta says is changing

Meta describes GEM as a recommendation-system foundation model that can transfer learning across advertising surfaces and objectives. The 2026 approach is designed to support goals from awareness and consideration through to conversions, using a shared modelling approach and expanded computing capacity.

Why advertisers should care

Ad delivery systems influence which people see an ad and how impressions are allocated. Improvements to ranking models could affect delivery and performance, but advertisers should not assume a platform-level model update will automatically improve every account or campaign.

Practical implications for marketing teams

  • Keep conversion signals reliable: Ensure event tracking and pixel or Conversions API configuration are accurate.
  • Test creative variety: Provide distinct, policy-compliant creative options that match the campaign objective.
  • Evaluate business outcomes: Track qualified leads, purchases, cost per acquisition and return on ad spend.
  • Use controlled comparisons: Consider seasonality, budgets, creative changes and auction competition before attributing results to a model update.

Meta’s engineering article describes its system design and ambitions; it does not guarantee a specific uplift for every advertiser or announce a separate rollout of new ad placements.

Source

Meta Engineering: Meta’s Generative Ads Model (GEM): Scaling the Central Model and Compute Engine for Frontier Ads AI (October 13, 2026)

Technical descriptions and performance claims are attributed to Meta.