Google AI Overviews did not arrive quietly. When Google began rolling them out broadly in May 2024, the SEO community's immediate concern was traffic loss. If Google answers the question on the results page, why would anyone click through to the source? That concern was legitimate but it framed the wrong problem. The more useful question was not how to stop AI Overviews from appearing. It was how to become the source they cite.

Understanding how AI Overviews select their sources is the foundation of AEO research. This article documents what is currently understood about the selection process, based on Google's public communications, observable citation patterns, and the emerging body of AEO research being built by practitioners tracking the system's behaviour since its rollout.

---

What are Google AI Overviews?

Google AI Overviews are AI-generated summaries that appear at the top of certain search results pages, above organic blue link results. They synthesise information from multiple web sources into a single coherent answer, with citations linking to the sources the system drew from. They appear for a subset of queries, primarily informational ones where a synthesised answer is more immediately useful than a list of links.

AI Overviews are powered by Google's Gemini large language model, integrated with Google's existing search index. The system does not browse the web in real time for each query. It draws from Google's indexed content, which means pages that are not indexed cannot be cited regardless of their quality. Indexing remains the prerequisite. Citation selection happens within the indexed pool.

The format varies depending on the query. Some AI Overviews are a single paragraph. Others include bulleted lists, numbered steps, or comparative tables. The format is determined by the query type and the structure of the source content the system identifies as most relevant.

---

How does Google decide which queries trigger an AI Overview?

Not every search query produces an AI Overview. Google's systems evaluate each query to determine whether a synthesised answer is more useful than a ranked list of links. Based on observable patterns, AI Overviews appear most consistently for:

• Informational queries. Questions beginning with what, how, why, when, and which are the most consistent triggers. These are queries where the user is seeking to understand something rather than to navigate to a specific site or complete a transaction.
• Multi-part questions. Queries that ask for a comparison, a sequence of steps, or an explanation of a relationship between concepts frequently produce AI Overviews because the synthesised format handles complexity better than a single ranked result.
• Queries with consensus answers. When multiple high-quality sources agree on the answer to a question, the AI system can synthesise that consensus with high confidence. Queries where sources disagree significantly are less likely to trigger an Overview or more likely to produce one with hedging language.

AI Overviews appear less frequently for navigational queries (searches for a specific brand or website), transactional queries (searches with clear purchase intent), and highly specialised professional queries where Google's systems are less confident about source quality.

---

What factors influence which pages get cited in an AI Overview?

This is the central question for AEO research and the one with the least settled answer. Google has not published a definitive list of citation ranking factors for AI Overviews. What is known comes from three sources: Google's general guidance on content quality, observable patterns in which pages get cited, and emerging research from practitioners tracking citation behaviour at scale.

The factors that appear most consistently associated with citation are:

Organic ranking position. The strongest single predictor of AI Overview citation is where a page ranks in standard organic results. Research published in early 2026 identified a 92 percent correlation between pages in Google's top 10 organic results and pages cited in AI Overviews for the same queries. Pages outside the top 10 are cited occasionally but rarely. This means the foundation of citation eligibility is standard SEO, not a separate optimisation track.

Content structure. Within the pool of organically ranking pages, content structure is the primary differentiator. Pages that state a direct answer in the first paragraph under a clearly labelled heading are cited more frequently than pages that present the same information in unstructured prose or delay the answer with lengthy introductions. AI systems need to extract a clean passage. The easier the extraction, the higher the probability of citation.

Topical authority. Pages on sites that cover a topic consistently and in depth are cited more frequently than pages on sites that cover many topics loosely. A page about dive computer calibration on a site dedicated entirely to scuba diving equipment carries more topical authority in that domain than the same information on a general outdoor sports blog. Google's systems appear to weight source authority at the domain level, not only at the page level.

Content freshness. Pages updated within the last 12 months account for the large majority of AI Overview citations. Stale content is deprioritised regardless of its original quality. This reflects the AI system's need to provide accurate, current information. A page about Google's algorithm that was last updated in 2023 is not a reliable source for a query about 2026 ranking behaviour.

Schema markup. Pages with structured data, particularly FAQPage, Article, and BreadcrumbList schema, are more legible to AI systems. Schema provides explicit signals about content type, content structure, and how the information on the page is organised. A page with FAQPage schema tells the system exactly where the questions and answers are located, which simplifies extraction.

Source trustworthiness. Pages on sites with clear author attribution, verifiable credentials, About pages with real team information, and external citations pointing to them are cited more frequently than anonymous or opaque sources covering the same information. This is the E-E-A-T dimension of AEO.

We have run early observations on this specific factor and the pattern is consistent enough to be worth noting. Sites that were missing one or more of these trust signals, no named author, no About page, no external references pointing to them, appeared in AI Overview citations at a noticeably lower rate than sites covering the same topic with equivalent content quality but complete trust signals in place. The difference was not dramatic in every case but it was directional and consistent across the sites we tracked.

The broader implication we think this points to is significant. If source trustworthiness becomes a harder requirement for AI citation rather than a soft advantage, it changes the viable format for content sites fundamentally. General information sites with no clear niche, no named authors, and no demonstrated expertise in a specific domain will find it increasingly difficult to earn citations regardless of how well their content is structured. The sites that benefit most from AEO are those that have gone deep on a specific topic with identifiable experts behind the content. We observed this pattern directly: sites covering a narrow niche with genuine expertise signals were cited. Sites covering broader topics with anonymous authorship were not, even when their content addressed the same query.

This is one observation set from a limited sample and we are not ready to publish it as a controlled experiment result. But it is consistent enough with what Google's public guidance suggests that we think it is worth flagging here. We will run a properly controlled test on this specific variable and publish the results in the AEO Research category when the data is ready.

---

How is AI Overview citation different from featured snippet selection?

Featured snippets and AI Overviews are related but distinct systems. Understanding the difference matters for how you optimise content for each.

A featured snippet pulls a single passage from a single page and displays it at the top of results. The source page is always visible and always linked. The user can see immediately where the information comes from and click through to read more.

An AI Overview synthesises information from multiple pages into a single generated answer. The citations are present but secondary, displayed as small reference numbers rather than as prominently linked sources. The generated text may not quote any single source directly but instead paraphrase, combine, or restructure information from several pages simultaneously.

The practical implication is that optimising for featured snippets and optimising for AI Overview citations are complementary but not identical. Featured snippet optimisation focuses on extractable single passages that directly answer a specific query. AI Overview optimisation focuses on being a reliable, well-structured, topically authoritative source that the system trusts enough to draw from, even if it synthesises your content rather than quoting it directly.

---

What happens to traffic when an AI Overview cites your page?

The traffic impact of AI Overview citation is one of the most actively researched questions in the AEO field right now and one of the least definitively answered. The honest position is that the data is still thin and the patterns are emerging rather than established.

What the available research suggests is that citation in an AI Overview does not consistently produce a measurable click-through traffic increase to the cited page. The user reads the synthesised answer and in many cases does not need to visit the source. This confirms the concern the SEO community raised at AI Overviews' launch.

However, two secondary effects appear to be real even when clicks do not materialise. First, brand visibility: a site name appearing consistently in AI Overview citations builds brand recognition with users who are in the research phase of a decision, even if they do not click immediately. Second, the users who do click through from AI citations tend to be higher-intent visitors, arriving with more context and more specific interest than average organic visitors. The conversion rate difference between these visitors and standard organic traffic is an active area of measurement.

At KatvTech we are tracking citation appearances and their associated traffic and behaviour patterns across our experiment sites. We will publish the data when we have a sample large enough to be meaningful. For now, the most defensible position is that AI Overview citation is a brand signal worth optimising for even in the absence of a reliable click-through uplift figure.

---

How do ChatGPT and Perplexity differ from Google AI Overviews in citation behaviour?

Google AI Overviews, ChatGPT, and Perplexity are three distinct systems with different underlying architectures and different approaches to source selection. Treating them as identical for AEO purposes is a common mistake.

Google AI Overviews draw from Google's own search index and weight sources using signals consistent with Google's organic ranking algorithm. A page that ranks well in Google organic search is significantly more likely to be cited in a Google AI Overview than a page that does not.

ChatGPT with web browsing enabled uses Bing's index as its primary source for real-time web content. This means pages that rank well in Bing, which uses different weighting than Google, may be cited in ChatGPT at different rates than pages that rank well in Google. Reddit is consistently among the most cited domains in ChatGPT responses, a pattern that reflects both its strong indexing across Bing and the community-generated, experience-rich nature of its content.

Perplexity operates as a dedicated answer engine with its own crawling and indexing system. Its citation behaviour appears to weight recency and factual specificity more heavily than domain authority, which means newer content from smaller sites can earn citations if it contains specific, verifiable information that the system identifies as relevant and reliable.

The content structure factors that improve citation probability are consistent across all three systems: direct answers, clear headings, specific facts, and schema markup. The domain authority and indexing factors vary by platform. Optimising content structure first and then building domain authority is the most platform-agnostic approach to AEO in 2026.

---

What should you do with this information?

The practical implications of how AI Overviews work can be distilled into a short list of actions that are worth taking on every piece of content you publish, regardless of whether an AI Overview currently appears for your target query. The queries that trigger AI Overviews change as Google refines the system, and content published now may become a citation source for queries that do not yet produce Overviews.

• Ensure every article is indexed. Unindexed pages cannot be cited. Check Google Search Console Coverage report regularly.
• State the direct answer to your article's main question in the first paragraph, not after an introduction.
• Use question-based H2 headings that match the language of real search queries.
• Add FAQPage schema to every article using Yoast's FAQ block. Five to eight questions per article targeting related queries.
• Update high-value content at least every six months. Add a visible last-updated date.
• Build topical depth on your domain before trying to earn citations on competitive queries. A site with 30 well-structured articles on a specific topic will earn citations before a site with 5 articles covering a broader range.

The AEO Research category on this site documents our ongoing experiments testing each of these factors in isolation. As those results accumulate, the picture of what actually moves citation probability, as distinct from what appears correlated with it, will become clearer.

---