The pipeline

  1. Trigger decision. Not every query gets an Overview. Trigger rates cluster around 16–25% on broad panels (higher on clickstream panels), with informational queries dominating (~88% of triggers) and commercial queries growing (from ~8% to ~18% of triggers during 2025). Long, question-shaped queries trigger most reliably.
  2. Query fan-out. The query expands into hidden sub-queries — "best CRM for a 20-person logistics company" fans into features, pricing, company-size fit and industry fit. Each sub-query retrieves its own candidate passages from the Google index.
  3. Grounding. Gemini (Gemini 3 globally since January 2026; Gemini 3.5 Flash after the I/O 2026 merger) drafts an answer constrained by the retrieved passages, with the Knowledge Graph supplying entity facts.
  4. Citation cards. Link cards appear beside the answer — typically a handful per Overview. They are the visible output of which passages grounded which sentences.

What the 2026 AI Mode merger changed

AI Mode — Google's conversational search surface, which passed a billion monthly users in May 2026 — merged with AI Overviews at I/O 2026 into "one seamless AI Search experience". Practically: the quick snapshot now hands off into conversation, carrying context. For brands this means an Overview citation is no longer a single impression — it seeds a session where follow-up questions keep pulling from the same grounded sources. Winning the first citation compounds inside the conversation.

Citation behavior

  • Index-dependent. Pages that don't rank meaningfully in classic search rarely appear in citation cards. Google has said its quality systems apply; independent analyses consistently find heavy overlap between top organic results and Overview sources — but not identity: quotable structure can pull a lower-ranking page into the cards.
  • Reference-weighted. YouTube, Wikipedia and Google's own properties are the most-cited source classes in broad analyses — the Knowledge Graph and its trusted-reference layer anchor entity answers.
  • Freshness-sensitive for volatile topics. Newsy queries pull recent pages; evergreen definitional queries tolerate older but authoritative pages.

Optimization checklist

  1. Hold classic rankings for target queries — the index is the retrieval base; technical SEO health is non-negotiable.
  2. Reformat top pages into answer-block structure — direct 40–80 word answers under question-shaped H2s; comparison tables for "X vs Y" queries.
  3. Feed the Knowledge Graph — Organization schema with sameAs, Wikidata entry where eligible, consistent entity facts everywhere.
  4. Cover the fan-out — for each money query, map the likely sub-questions (features, pricing, fit, trust) and ensure a passage answers each.
  5. Mind the click economics — position-1 CTR falls ~58% under an Overview (Ahrefs, Feb 2026); measure citations, not just clicks.

Measurement notes

Track Overview presence for your prompt set (many queries still don't trigger one), your appearance in citation cards, and whether follow-up conversational context retains your brand. Google's documentation for AI features in Search and the Gemini grounding API are the primary sources; trigger-rate panels (Semrush, Similarweb, BrightEdge) disagree by methodology — read them as ranges.

Sources: Alphabet Q2 2025–Q2 2026 earnings calls; Google I/O 2025/2026 announcements; Ahrefs CTR studies (Apr 2025, Feb 2026); Pew Research (Jul 2025); Semrush trigger-rate tracking (2025). Figures are dated where cited; panels differ in methodology.

Frequently asked questions

Does ranking #1 guarantee an AI Overview citation?
No — overlap is heavy but not identity. Overview citation adds a passage-quality filter: quotable, self-contained chunks can pull a lower-ranking page into the citation cards, while a thin top-ranking page can be skipped. Rankings get you into the candidate pool; passage structure wins the card.
What is query fan-out in practice?
One question becomes many hidden sub-queries, each retrieving its own passages. 'Best CRM for a 20-person logistics company' fans into feature, pricing, company-size and industry sub-queries. Your content must win passages across several of them — covering only the head term wins a fraction of the answer.
Is AI Mode replacing classic Google search?
It is absorbing it. The 2026 merger made AI Overviews and AI Mode one experience, with the snapshot handing off into conversation. Classic ten-blue-links persist underneath, but the answer layer is now the default surface for a growing share of queries.