Google Discover Expands Personalization For Publishers

A January 2026 analysis in Search Engine Journal outlines how Google Discover uses large-scale recommender techniques derived from YouTube’s two-tower architecture. It details candidate generation, item/user embeddings, freshness-vs-relevance tradeoffs, and hybrid signals (search, YouTube, Knowledge Graph), and notes publisher tactics such as regular posting, large images, mobile-first design and Search Console monitoring. The piece highlights implications for publishers seeking stable traffic from Discover's ML-driven pipeline.