What an AI Overview actually is
An AI Overview is the AI-generated answer block that Google shows at the top of certain search results pages. It sits above the regular organic results, sometimes above the ads, and presents a synthesised answer drawn from a small set of cited sources. Each cited source appears as a small linked tile underneath the answer.
Two important framings to start with. First: AI Overviews are part of the Google SERP, not a separate product. The user does not opt in. They search for something, and if Google's systems decide an AI Overview is appropriate for that query, the Overview appears at the top. Second: AI Overviews are generated dynamically per query. The same query run twice can produce slightly different Overviews and slightly different citation sets, particularly during the first few weeks after the topic emerges in the index.
The Overview block typically contains a short paragraph of synthesised answer, occasionally a bullet list, sometimes a comparison table or a short series of sub-headings. Below the answer text is the citation row, usually three to five linked sources. Users can expand the Overview to see longer answers and more citations, but most do not.
The first cluster in this pillar, AEO vs GEO vs SEO, situates AI Overviews inside the wider AI search picture. This chapter goes deeper specifically on AI Overviews because they touch more Australian businesses than any other AI search surface.
How they get generated
The full algorithm is not public. What Google has confirmed plus what we observe across client sites suggests a pipeline something like the following. None of this is algorithmic certainty.
- Query classification. Google's systems decide whether the query is suitable for an AI Overview. Informational and how-to queries are suitable; pure navigational queries usually are not. The classification appears to be probabilistic rather than fixed; the same query may or may not trigger an Overview depending on the context.
- Candidate source selection. The systems pull a candidate set of sources, drawn heavily from the top organic ranks for the query plus selected high-authority sources Google's systems already trust on the topic.
- Synthesis. An LLM (the Gemini family inside Google's stack) reads the candidate sources and produces a synthesised answer. The synthesis prefers content that directly addresses the query, has clean structure, and contains citable claims.
- Citation selection. The synthesis step selects a small number of the candidate sources to display as citation tiles. The selection appears to favour sources whose passages were most directly used in the synthesised answer.
- Render. The Overview gets rendered at the top of the SERP, above the regular organic results. The user sees the answer and the citations.
The implication for SEO. The candidate set is drawn from the existing top organic results, which means traditional ranking is still the entry ticket. The citation choice within that set is influenced by how lift-able the page's content is, which is where the AI-specific moves come in.
What triggers an AI Overview
Six observations from monitoring AI Overview coverage across client query sets through 2025 and 2026. None are algorithmic facts; all are observed patterns.
- Informational and how-to queries trigger Overviews most often. "How does X work", "what is Y", "best way to do Z". Coverage on these query types is high.
- Brand queries rarely trigger Overviews. Searching for a specific company name almost always returns the regular SERP with the brand's own pages and Knowledge Panel, no Overview.
- Transactional queries are inconsistent. Some product-comparison and purchase-intent queries trigger Overviews, particularly comparison-style ones. Many do not.
- Local-intent queries rarely trigger Overviews. "Plumber near me", "best cafe in Fremantle". Local pack and Maps still dominate these.
- YMYL queries trigger Overviews but with conservative source selection. Health, finance and legal queries often show an Overview, but the cited sources skew heavily toward Wikipedia, government sites, established institutions.
- Long-tail queries trigger Overviews more often than head terms. The longer and more specific the query, the more likely an Overview appears. Head terms with strong commercial intent are less likely to show one.
The practical implication for an Australian business: check which of your query set actually triggers Overviews before assuming you need to rebuild everything. A Perth plumbing site competing for "plumber Fremantle" and "blocked drain Joondalup" has very little AI Overview exposure. A health publisher competing for "symptoms of X" has very high exposure. The right response depends on the mix.
The click-volume impact
The most-asked question. Our working position based on what we measure across client GA4 properties and GSC accounts:
- Pure informational queries. 20 to 30 percent reduction in click volume. The user gets a usable answer in the Overview and never clicks. This is the headline number that gets quoted in industry articles, and it is real for the queries it applies to.
- Commercial-intent queries. Less than 10 percent reduction in click volume on most. The user still needs to evaluate vendors, compare options, or initiate a transaction. The Overview gives context; the click is still required.
- Local-intent queries. Minimal impact. The local pack and Maps still dominate the SERP. Where an Overview appears, it usually sits above the local pack rather than replacing it, and click-through to local listings stays similar.
- Brand queries. No measurable impact. Brand queries rarely trigger Overviews and brand clicks are not affected.
The aggregate impact across a typical client's site depends on the query mix. A B2B content publisher with mostly informational traffic might see a meaningful aggregate drop. A trade business with mostly local commercial intent might see no measurable change. The lesson: do not panic until you have measured your own mix.
One more wrinkle. The clicks the Overview takes are usually the lower-intent informational clicks anyway. The clicks that survive are often the higher-intent commercial and comparison clicks. Net revenue impact is sometimes less than net click impact, because the surviving clicks convert better.
How pages get cited
The pattern across the pages we see cited is consistent. Five traits show up repeatedly.
Trait 1: Already ranking near the top
Pages cited in AI Overviews are almost always already ranking in the top 3 to 5 organic positions on the underlying SERP. The Overview citation set is drawn from the top of the regular results plus a small number of high-authority sources Google already trusts. Below position 5, citation odds drop sharply. The implication: if your page is not on page one, win the ranking first before chasing the citation. Standard SEO from the On-Page SEO pillar and Content Strategy pillar still does most of this work.
Trait 2: Clean entity identity
Cited pages tend to come from sites with locked-down entity identity. Organization schema with sameAs links to brand profiles, Person schema for named authors, consistent NAP across the site, transparent About page. Sites where the identity is ambiguous (no schema, no named authors, generic About) get summarised but rarely cited. See entity SEO.
Trait 3: A lift-able direct answer near the top
Pages where the answer to the underlying query is stated clearly in the first 50 to 100 words, in a single declarative passage, get cited more often than pages where the answer is buried lower or split across multiple paragraphs. The pattern is the same as for traditional featured snippets, just sharpened. See featured snippets for the page-level treatment that carries over.
Trait 4: Strong E-E-A-T signals
Named credentialled authors. Citable claims with named sources. A transparent organisational identity. Original information the AI cannot regurgitate from elsewhere. On YMYL topics this trait moves from "helpful" to "required". See E-E-A-T explained.
Trait 5: Clean structured data
Valid FAQPage schema on FAQ blocks. Valid HowTo schema on step-by-step content. Valid Article schema with proper author and publisher fields. The structured data gives the AI system a confidence boost about what the page is and which entity to attribute it to. See schema for AI.
Step-by-step optimisation
The order matters. Skip ahead and the moves cost effort without payoff.
Step 1: Check whether your queries trigger Overviews
Pull your top 20 to 50 priority queries from GSC. Run each one manually in Google Search. Record which ones return an AI Overview, which sources are cited, and which of the cited sources is the closest competitor to you. If your queries do not trigger Overviews at all, this whole chapter is academic for your site.
Step 2: Confirm the target page is in the top 10 organic results
Citations come almost exclusively from the top of the existing SERP. If the target page is on page two or lower, the project is to win the ranking first via content depth, topical authority and the off-page work. Chasing the AI Overview citation while ranked at position 25 is the wrong order.
Step 3: Tighten the entity layer
Add valid Organization schema with sameAs links. Add Person schema for each author with a real bio page. Add LocalBusiness if relevant. Confirm the schema validates in Google's Rich Results Test. This is the single move with the highest payoff across both AI Overviews and the other AI clients.
Step 4: Surface the direct answer near the top
Rework the opening of the page so the answer to the underlying query is stated clearly in the first 50 to 100 words, as a single declarative passage. Not buried under a story, not split across three paragraphs. The AI extractor lifts this kind of passage cleanly; meandering introductions do not survive the extraction.
Step 5: Add FAQ block with FAQPage schema
Three to six question-and-answer pairs at the end of the page that cover the natural follow-up questions a user might ask. Wrap them in valid FAQPage schema. Use the actual question as the summary, not a marketing-style rephrasing.
Step 6: Track citation pickup monthly
Re-run the target queries monthly. Record which sources are cited. Note when your page gets picked up. Watch for citation churn (you are cited one month, not the next) and look for the reason. AI Overview citation behaviour changes more abruptly than traditional rankings, so the monthly check is non-negotiable.
Common mistakes
- Measuring your own query mix before assuming the 20 to 30 percent click loss applies to you.
- Winning the underlying ranking first, then optimising for the citation.
- Tightening entity identity with
OrganizationandPersonschema. - Surfacing the direct answer in the first 50 to 100 words.
- Tracking citation pickup monthly so you catch churn early.
- Panicking at the headline 20 to 30 percent click loss without checking which queries it applies to.
- Chasing citations on pages that are not in the top 10 organic positions.
- Ignoring the entity layer because "Google works it out". AI Overviews are less forgiving.
- Burying the answer four paragraphs in. The extractor lifts the top, not the middle.
- Assuming AI Overview citation behaviour is stable enough to set and forget.
Perth and WA context
Three patterns we see for Perth and WA businesses specifically.
Local trade businesses are barely affected. "Plumber Fremantle", "electrician Joondalup", "tree lopping Mandurah". These queries almost never show an AI Overview. The local pack and Maps still dominate. The work that matters for these businesses is the Local SEO pillar plus the Google Business Profile chapter, not AI Overview optimisation. See Local SEO Perth.
Service businesses with informational content blogs are mixed. A Perth law firm or healthcare practice with an informational blog ("what is X procedure", "your rights when Y happens") sees AI Overviews on the blog queries and not on the commercial service queries. The right move is to invest in AI Overview optimisation on the informational layer (entity identity, FAQ schema, direct answers) while leaving the commercial pages focused on conversion. See legal SEO and healthcare SEO.
WA mining and resources content is high-exposure on top-of-funnel content. The educational content used to capture early-stage interest ("what is a planned shutdown", "FIFO roster types", "mining equipment classifications") is squarely in the AI Overview triggering zone. The right move is to either invest properly in AI Overview optimisation for that layer or accept the click reduction and refocus the budget on mid-funnel commercial content. See mining SEO and SEO Karratha.
For the wider context, the AI Search pillar covers the full visibility stack. The search intent chapter covers how to read the intent of a query before deciding whether it is an AI Overview risk. The featured snippets chapter covers the precursor optimisation pattern that carries over almost completely. For a unified view of where your site sits across SEO and AI Overview readiness, the free SEO audit includes the relevant signals, and the website audit service goes deeper.