Query Fan-Out.

Classic search is a one-to-one relationship: one query, one list of results. Query fan-out breaks that apart. The language model reads the input, identifies entities, constraints and the actual intent behind it, and derives several searches which it then runs simultaneously. 'Which agency does SEO and Google Ads for small companies in Zug' becomes sub-queries about SEO agencies in Zug, about Google Ads support for SMEs, about pricing and about references. Only the merged result becomes the answer shown in AI Mode or in AI Overviews.

For visibility this has an uncomfortable consequence. A page can rank very well for the original phrasing and still never be cited, because it answers none of the derived sub-questions cleanly. The reverse happens too: a page that never ranks for the head term can appear as evidence for a single sub-question. That is why citations in AI answers and positions in the results list routinely diverge. Working for fan-out therefore means thinking in bundles of questions around a topic, not in individual keywords.

Those bundles can be reconstructed rather than guessed. Three sources are enough: the fully phrased questions in Google Search Console, the 'People also ask' suggestions on the results page, and the questions clients actually raise in a first conversation. Together they produce a list of ten to twenty sub-questions per topic. Each gets its own heading on the page, with an answer underneath that makes sense on its own. Those self-contained sections are exactly the units the fan-out collects.

A practical example: a dental practice in Küsnacht had a well-ranking page about implants. It never appeared in AI answers, because the derived sub-questions about cost, duration, pain, health insurance and aftercare were only mentioned in passing. After restructuring the page into five clearly headed sections, each with a direct answer, the ranking stayed the same but the page became citable for several sub-questions. The work sat in the structure, not in additional text.