How to show up in Google AI Overviews

AI Overviews now sit at the top of many Google results, answering the question before a single blue link. Getting cited there is a different game from ranking #1 - here's how to play it.

Google's AI Overviews summarise an answer at the top of the results page and cite a handful of sources. For many queries they push the classic links below the fold - so being cited in the Overview is increasingly where the visibility (and clicks) are. The good news: the fundamentals overlap heavily with good SEO. The catch: ranking #1 no longer guarantees you're in the Overview.

How AI Overviews choose sources

Overviews are generated by Google's models using a technique often described as "query fan-out": your query is split into several sub-questions, each searched, and the results synthesised. The sources cited tend to be pages that directly and concisely answer a specific sub-question, are trustworthy, and align with what other sources say. So the target isn't one fat keyword page - it's clear answers to the many small questions buried inside the main one.

Curious how AI engines describe your brand right now? Get a free visibility audit and see where you stand across ChatGPT, Gemini and Perplexity.

Step 1: Answer sub-questions explicitly

Break the big topic into the specific questions people actually ask, and answer each one in a self-contained passage - ideally a clear heading followed by a direct, 2-4 sentence answer. Overviews love passages that can be lifted whole without needing the rest of the page for context.

Step 2: Lead with the answer

Put the conclusion first, then the reasoning. Inverted-pyramid writing - answer, then support - matches how the model extracts and quotes. Burying the answer under 600 words of preamble makes your page harder to cite.

Step 3: Earn trust signals

Overviews favour sources Google already trusts. That means the usual E-E-A-T groundwork: clear authorship, accurate and current facts, citations to primary sources, and a track record on the topic. Fresh, well-maintained content beats stale pages - keep dateModified honest and update material that ages.

Step 4: Use structured data and clean formatting

Headings, short paragraphs, lists, tables and FAQ schema all make it easier for the model to identify and extract a discrete answer. Structured, scannable pages are simply easier to quote than walls of text.

Step 5: Cover the topic, not just the keyword

Because fan-out pulls from many sub-queries, topical breadth helps. A cluster of pages that thoroughly covers a subject - definitions, comparisons, how-tos, edge cases - gives the model more chances to cite you across the sub-questions than a single page ever could.

Step 6: Track your presence

AI Overview citations aren't in standard rank trackers, and they're volatile - here this week, gone next. Monitor which queries trigger an Overview in your space and whether you're cited, so you can spot losses and defend your place. This is the same monitoring discipline AEO platforms apply across ChatGPT, Perplexity and the rest.

"A page can rank well and still get skipped because the answer is spread across four paragraphs instead of sitting in one liftable block."

Write for the passage, not just the page

AI Overviews rarely quote a whole page. They lift a passage - usually a heading and the two or three sentences beneath it - and drop it into the summary with a citation chip. That changes what "good content" means. A page can rank well and still get skipped because the answer is spread across four paragraphs instead of sitting in one liftable block.

The fix is mechanical. Make each answer survive being copied out on its own:

• Front-load the specific claim. "Standard shipping takes 3-5 business days" beats "Shipping times vary depending on a number of factors."
• Resolve pronouns and context inside the passage. If the sentence only makes sense after reading the paragraph above it, the model has to work harder to extract it - and may pick a competitor whose passage stands alone.
• Keep the answer block short. A 40-60 word answer under a question-shaped heading is easier to quote than a 300-word essay. Add the depth after the direct answer, not before it.
• Match the heading to the sub-question. Phrase headings the way a person asks, not the way a marketer titles. "How much does X cost?" gets matched to a fan-out sub-query more reliably than "Pricing."

A quick test: read only your heading and the first two sentences under it. If that alone answers the question, you have a passage worth citing. If it doesn't, rewrite until it does.

Where schema actually helps (and where it doesn't)

Structured data is not a ranking cheat code for Overviews, and anyone promising that is overselling it. Google generates the summary from the content it reads, not from your JSON-LD. What schema does is remove ambiguity - it tells Google exactly what a passage is, which makes clean extraction more likely and eligibility for related surfaces more reliable.

The schema types worth the effort for AEO:

• FAQPage - pairs a question with a self-contained answer, which mirrors the sub-question structure fan-out is looking for. Only mark up questions that genuinely appear on the page.
• HowTo - useful for procedural queries where the Overview may summarise steps.
• Article with author and dateModified - reinforces the authorship and freshness signals that feed trust. Keep dateModified honest; faking recency is a good way to lose the trust you are trying to build.
• Product, Organization and breadcrumb - help Google understand entity and context, which matters when the Overview needs to disambiguate who you are.

What schema won't do: rescue a vague answer, invent authority you haven't earned, or force a citation. Think of it as making an already-good passage machine-legible, not as a substitute for the passage being good.

How AI Overviews differ from AI Mode - and why you optimise for both

People conflate these two, but they behave differently and reward slightly different things. AI Overviews appear automatically at the top of a normal results page for informational queries, sit alongside the classic blue links, and summarise in a few sentences with a handful of citations. AI Mode is a separate, conversational surface a user actively switches into for complex, multi-step research - and, as reported, it does not show organic results at all, leaning much harder on fan-out to run many sub-queries in parallel before answering.

The practical implications:

• AI Mode punishes thin coverage more. Because it fans out aggressively and has no blue links to fall back on, breadth across a topic cluster matters even more than it does for Overviews. One page answering one question is easy to miss.
• Conversational phrasing helps in AI Mode. Follow-up questions ("what about for teams?", "how does that compare to X?") map to content that anticipates the next question, not just the first one.
• The good news: the fundamentals overlap. Self-contained passages, clear answers, earned trust and topical depth serve both surfaces. You are not maintaining two content strategies - you are making one strategy robust enough to survive fan-out.

Google reported AI Overviews crossed 2.5 billion monthly users in 2026, so this is not a niche surface to hedge against. Optimise for the Overview first, then confirm your cluster is deep enough that AI Mode can't route around you.

What to actually measure

"Are we in the Overview?" is a yes/no that hides most of the signal. Track it at the query level, over time, because Overview citations are volatile - present one week, gone the next - and a single snapshot tells you almost nothing. Build a small dashboard around these metrics:

• Overview trigger rate - what share of your priority queries even produce an Overview. If a query doesn't trigger one, classic SEO still owns it.
• Citation share - of the queries that do trigger, how often you are among the cited sources, and who shows up when you don't.
• Passage attribution - which specific page and passage got cited, so you can see what "good" looks like on your own site and replicate the pattern.
• Volatility - how often your citation appears and disappears for the same query. High churn tells you the position is contested and worth defending.
• Downstream clicks - Overviews can suppress clicks even when you're cited, so watch impressions-to-clicks on affected queries rather than assuming a citation equals traffic.

None of this lives in a standard rank tracker, which is why AEO platforms exist - to log citations across Overviews, ChatGPT, Perplexity, Gemini and Copilot, then alert you when you lose a spot you used to hold. Whether you buy a tool or build a scrappy manual check, the discipline is the same: measure citation at the query level, on a schedule, and treat a lost citation like a lost ranking.

The bottom line

Optimising for AI Overviews is SEO with the emphasis moved: from ranking a page to being the clearest, most trusted answer to a specific question. Answer sub-questions directly, lead with the answer, earn trust, structure the page, and cover the topic in depth - then watch whether it's working.

→ Related: Query fan-out, how AI turns one question into many searches