Julian broke down Microsoft's insider view of how AI search actually selects its sources — straight from an SEO Week 2026 keynote by a Microsoft AI Product Manager. The big idea: AI doesn't rank pages, it runs a multi-step evidence construction process across dozens of parallel sub-queries, and only sources with broad topical coverage survive it.
1. 🔀 Query Fanout: AI decomposes one question into 5–100+ parallel sub-queries targeting different topical angles — not keyword variations.
2. 🏷️ Entity Anchoring: Fan-out queries resolve to named, established entities, giving brands with strong entity presence a structural advantage in retrieval.
3. 🧠 Multi-Model Judgment: Before an answer is assembled, multiple models cross-check candidate sources in a consensus layer that's invisible from the outside.
4. 📊 Citation Share vs. Recommendation: Citation share audits topical authority, but brand recommendation in commercial queries — driven by reviews and third-party mentions — is the real revenue metric.
5. 🗺️ Semantic Breadth Wins: Cover the full entity landscape around your topic; referencing competitors and adjacent platforms signals market context and increases topical credibility.
6. 🔧 Technical Foundations: Confirm AI bot crawl access, submit via IndexNow for Bing/ChatGPT visibility, and publish full transcripts so video and podcast content is machine-readable.
Julian will be out for three weeks — pre-published guides on schema markup, AI literacy, and AI frameworks will keep the community building in the meantime.
Thank you for making us a #1 bestseller! The revolution is officially here.