Getting cited in Google AI Overviews and AI Mode.

Google AI Overviews cites sources drawn from the regular Google search index, so eligibility starts with the same basics as classical SEO. A page needs to be indexed without a noindex tag, needs to open with a self-contained answer paragraph within the first 100 words rather than burying the answer further down, and should use question-shaped sub-headings that match how a real user actually phrases that question when searching. On top of that, the page needs Article and Author schema marked up correctly, and any factual claim should cite a primary source directly. Google AI Mode grounds on the same underlying index, so the same fundamentals apply there too.

AI Mode does not use a different index than regular Google search. Both AI Mode and AI Overviews ground their answers on Google's main search index, the same one classical search results draw from. The real difference between the two is the surface each one appears on: AI Overviews shows up inside the classical search results page, sitting above the usual list of blue links, while AI Mode is a separate, standalone conversational interface a user opens directly. Both surfaces use Gemini to compose the written answer and cite sources back to that same set of already-indexed pages.

Query fan-out describes how Gemini breaks a single question a user typed into multiple smaller sub-questions behind the scenes, then assembles its final answer out of passages that each match one of those sub-questions. A page that only answers the original headline question loses out to a page that has anticipated that fan-out and answers several of the sub-questions as well. Running a manual fan-out audit on a priority page, meaning listing out every sub-question a buyer would naturally ask next and making sure each one has its own dedicated heading on the page, is the single highest-leverage AI Overviews fix available to most SMBs.

Opting out of Google-Extended does not affect AI Overviews citations. Google-Extended is specifically the opt-out token that controls whether Gemini can use a site's content to train future models, not whether that content can be cited in a live answer today. AI Overviews and AI Mode both ground their answers on the regular Google search index rather than on model training data, so disallowing Google-Extended in a site's robots.txt file has no effect on citation eligibility. It only affects whether the content can be used as training data for a future version of Gemini.

AI Overviews weights two things heavily: pages that directly answer the exact question a user asked, and pages that show genuine first-hand experience rather than a generic rewrite of common knowledge. A typical Reddit or Quora thread does both of those things naturally, since the question becomes the page title, the top-voted comment becomes the answer, and the voice throughout is unmistakably that of a real person describing their own experience. SMBs that want to compete against that kind of content have to write pages that answer one specific question per page with comparable directness, and back up any claim with real, specific implementation details rather than vague generalities.

Most AI Overviews cite somewhere between three and eight sources per answer, with the exact number depending on how complex the underlying query is. A page competing for a single, narrow sub-query has a realistic shot at being one of those sources even without a top-three ranking for the broader head term. This is exactly why fan-out coverage matters more than headline-keyword ranking for AI Overviews specifically: a page can lose the competition for the main, broad query and still get cited for answering one of its narrower sub-questions well.

FAQPage schema helps AI Mode and other AI surfaces extract clean, structured question-and-answer pairs from a page, which makes that content easier for the model to quote accurately. It no longer triggers a separate FAQ rich result inside classical Google search results, since Google fully deprecated that specific rich result on May 7, 2026. The practical move is to keep FAQPage schema in place for the extraction benefit it still provides to AI engines, and to pair it with visible, readable on-page question-and-answer text, while no longer expecting a dedicated FAQ box to appear directly on the search results page itself.

A page can rank on page one of Google and still be missing from the AI Overview above it, and this happens often enough that it should be checked for deliberately. Ranking and citation are two genuinely separate evaluations that Google runs using different criteria. A page can hold a strong classical ranking position while a competitor's page wins the actual AI Overview citation instead, simply because that competing page answers the user's underlying sub-question more directly or more specifically. Auditing the AI Overview block separately from the plain blue-link ranking is now a necessary step for any SMB serious about AI visibility, not an optional extra check.