Three months of feeds reveal how trends, news, video, and ads flow through pipelines, and why some content gets broadcast-level visibility.
Read →and why it’s twice as powerful in France
Read →Why did your article get picked up, and by which part of the machine? Tracing 42 million cards, we show which pipeline selects what in the French feed, and how to play each one.
Read →Planning to apply your French Discover playbook to the English feed? Bad idea: YouTube video, AI Overviews and ads rewrite the rules there, measured across 42 million cards.
Read →Can a publisher run one Discover playbook in both markets? No: moonstone reaches twice as far in France, while the YouTube video cascade and AI Overviews stay English-only.
Read →Special edition: what our data reveals about e-commerce sites showing up on Discover
Read →Discover does not run on a single algorithm. Related, followed creators, high affinity, item-item collaborative filtering, prompt tuning: a tour of the engines competing for every card in your feed.
Read →Why does your feed swing from uncannily relevant to completely off topic in just two swipes? The answer lies in the facets Discover pulls from the article you just clicked.
Read →Google Discover (End of 2025): desktop, social networks, AI Overviews, personalization... and tightening security
Read →Daily Hub tried to merge embeddings, entities, and real-time context – and buckled under the system’s complexity.
Read →Even during vacation, Damien continues his research.
Read →A deep dive into the unseen systems behind Google Search – uncovering live experiments, entity-based infrastructure, AI agents, and more.
Read →