Topic
4 videos.
Ask an AI search tool one question and it may break it into several smaller searches before it writes an answer. Those are fan-out queries. Google has described this for AI Mode, though how any tool does it in detail is not public. The idea matters because a page can answer part of a question without matching the exact words typed.
All the sources are software companies or their events, and most talks report what a speaker found or infers from outside. A first look is Surfer's short explainer, from 2025, based on a Surfer study of typed queries. Krishna Madhavan of Microsoft follows the steps inside Bing in the AI Engine Pipeline, which is advanced. Nik Ranger's talk on a travel platform and Mike King's on relevance engineering are also from a 2026 conference and assume some background.
Sells in this field
Mike King argues that search now works on meaning while many SEO tools still lean on BM25 and TF-IDF, and goes on to relevance engineering and fan-out queries.
Sells in this field
Fan-out queries are at the centre of Nik Ranger's AI visibility study for a travel platform, where seven million synthetic queries produced 110 Google citations and just 2 from ChatGPT.
Sells in this field
Krishna Madhavan traces the steps from a search query to an AI answer on Bing, including the index, annotation, fan-out queries and the engineering choices that affect which brands get named.
Sells in this field
Why might a brand miss out on AI mentions even when it ranks first on Google? Surfer's study of over 1,600 queries says AI tools quietly expand typed queries into hidden fan-out queries.
