When you submit a query that triggers AI Mode, Google does not just look up the top 10 results. It performs query fan-out — generating multiple related sub-queries, retrieving passages (not full pages) from the index, then synthesizing an answer with inline citations.
The unit of retrieval is a passage. This is why a perfectly-structured paragraph buried on page 3 of your site can show up as the cited source while your homepage does not. Passage-level optimization is the new game.
Google AI Mode breaks every user query into 5-15 related sub-queries before answering. To win, your content needs to address not just the head query but the predictable fan-out.
Practical implication: hub-and-spoke content architecture is now mandatory. A pillar page that answers the head query, supported by 8-15 sub-pages that answer the predictable sub-queries, dramatically outperforms a single long page.
No. We have measured many cases where the cited source ranks #5-#15 organically because its passage was a better answer. Position helps but is not sufficient — passage-level structure matters more.
Yes — Google supports a `nosnippet` directive and the broader Google-Extended user-agent block. But opting out generally costs you traffic; we rarely recommend it for service businesses.
AI Overview eligibility shifts with every algorithm update. Most drops trace to a quality update, a freshness signal, or a competitor publishing a better passage. Re-audit the page against the 14-point checklist.
Yes — but unevenly. Pure-informational queries see 30-70% CTR drops. Commercial-intent queries are largely unaffected so far because users still want to compare and click. Restructure your content mix toward commercial intent.