AI Overviews Now Appear on 48% of Queries

ClickRadius Institute · Published May 26, 2026

In early 2026, AI Overviews — the synthesized answer boxes Google places above its results — appeared on roughly 15% of queries. By late May 2026, industry tracking data puts that figure at approximately 48%. Nearly one in two Google searches now begins with an AI-composed answer rather than a list of links, and the expansion happened in months, not years. This article examines what drove the threefold growth, which categories of queries are most affected, what the measured impact on click behavior looks like, and — most usefully — what determines which sources get cited inside an Overview versus buried beneath one.

The trajectory: 15% to 48% in months

When Google launched AI Overviews, coverage was cautious: mostly long-tail informational queries where a synthesized answer was clearly helpful and commercially low-stakes. Through 2025 the footprint expanded steadily, and in early 2026 industry trackers measured Overviews on roughly 15% of queries.

Then came the acceleration. In the run-up to and aftermath of Google I/O 2026 — where AI Mode became the default search experience and Sundar Pichai described the release as Google's "biggest upgrade to Search ever" — Overview coverage roughly tripled. Industry tracking data now places AI Overviews on approximately 48% of queries.

The direction of travel matters as much as the number. Google's VP of Search Elizabeth Reid framed the underlying shift at I/O 2026:

This is the biggest upgrade to our Search box in over 25 years.

— Elizabeth Reid, VP of Search, Google, at Google I/O 2026

A company does not describe a feature that way and then retreat from it. The 48% figure is a waypoint, not a ceiling: with AI Mode as the default surface, the synthesized answer is becoming the norm for the queries that remain outside full AI Mode as well.

Which queries trigger Overviews — and which don't

Coverage is not uniform. Based on observed patterns across query categories, Overviews concentrate where synthesis adds value:

High Overview coverage

• Informational questions — "how does X work," "what is the difference between A and B," symptoms, definitions, processes. These were the original Overview territory and remain near-saturated.
• Comparison and evaluation queries — "best CRM for small law firms," "X vs Y." The model builds the comparison table users used to assemble from five tabs.
• Local service questions — "do I need a permit to replace a water heater in Arizona." Increasingly answered directly, with providers named.

Lower Overview coverage

• Navigational queries — users looking for a specific site still get the site.
• Transactional endpoints — checkout-intent queries where the click is the point.
• Volatile or sensitive topics — areas where Google remains conservative about synthesized answers.

The strategic implication: the informational and comparison content that most businesses built their organic strategy around is precisely the content whose clicks Overviews absorb most aggressively.

The measured impact on clicks

Three statistics describe what the 48% expansion has done to the traffic economy:

• Zero-click searches have risen to roughly 60% of all queries, up from about 45% before the AI-answer era, per industry estimates.
• The #1 organic position's click-through rate has fallen from roughly 27% to roughly 11% — a decline of more than half, concentrated on queries where an Overview sits above the ranking.
• Inside full AI Mode, industry data puts zero-click behavior near 93% — a preview of where Overview-covered queries trend as users acclimate to reading answers rather than scanning links.

According to Google's public positioning, Overviews are designed to send "more qualified" clicks — users who click through after reading the synthesis arrive better-informed and closer to action. There is truth in that framing: cited sources report that the clicks they do receive behave more like referrals than cold traffic. But the arithmetic is unforgiving for sites that are merely ranked and never cited: they inherit the traffic decline without the citation exposure.

Cited versus buried: the new dividing line

An AI Overview creates two classes of content. Cited sources appear inside the answer itself — linked, named, and implicitly endorsed as the basis of what Google just told the user. Buried sources rank beneath the Overview, in a zone a shrinking minority of users ever reach.

What separates them? Two evidence bases converge on an answer.

The content signals: Princeton's GEO research

According to "GEO: Generative Engine Optimization," the Princeton study presented at KDD 2024, three on-page signals measurably raise the probability that generative engines cite a page:

1. Statistics — specific, verifiable numbers a model can quote with confidence;
2. Quotations — attributed statements from identifiable people or organizations;
3. Source citations — the page itself grounding its claims in credible references.

The common thread is verifiability. A generative engine composing an answer it must stand behind prefers source material that is already structured like evidence. ClickRadius's 6-category AI-readiness score weights these three signals directly because they are the citation levers with peer-reviewed support.

The entity signals: who you are, not just what you wrote

According to industry data, on-page optimization is now the foundation rather than the majority of the outcome: most of what drives AI citations is off-site — the consistency and breadth of your entity's presence across directories, databases, review platforms, and third-party coverage. Overviews cite entities the underlying models recognize as authoritative on the topic. Two pages of equal quality do not have equal citation odds if one belongs to an entity the model has encountered across dozens of trusted surfaces and the other belongs to a digital ghost.

Quotations, statistics, and citations aren't decoration — they are the three content signals with published evidence of increasing generative-engine citations.

— ClickRadius Institute, summarizing Princeton's KDD 2024 GEO findings

How to respond: a five-step program

1. Identify your Overview exposure. Run your most valuable queries and record: does an Overview appear, and is anyone in your industry cited in it? For most sectors, industry estimates suggest a large majority of brands have zero AI-search mentions — meaning the citation slots for your topics may still be effectively unclaimed.
2. Restructure the pages that answer those queries. Question-form headings, a direct one-to-three-sentence answer immediately below each, then depth. Add FAQPage and Article schema so the extraction is unambiguous.
3. Install evidence. Add real statistics with attribution, quotations with named sources, and citations to authoritative references on every page you need cited — the Princeton triad.
4. Build the entity. Consistent name/address/phone/description everywhere, complete directory and data-source profiles, structured data declaring your organization, and third-party coverage. This is slower than on-page work and, per industry data, drives the larger share of the outcome.
5. Monitor citations as a KPI. Overview citations shift as models and indexes update. Track whether you are cited — across Google and the other engines answering the same questions (ChatGPT, Perplexity, Claude, Grok) — the way you once tracked rankings.

Anatomy of an Overview citation: what "winning" looks like

It helps to be concrete about what an AI Overview citation actually is, because the term gets used loosely. A typical Overview on a commercial-informational query contains three layers:

1. The synthesized body — several paragraphs or a structured list composed by the model. Most of the source material that informed it is invisible here; the model consulted more than it credits.
2. Inline and side citations — a subset of consulted sources, surfaced as links attached to specific claims or displayed alongside the answer. These are the slots businesses compete for.
3. Named entities — businesses, products, or organizations mentioned by name in the body text itself, with or without a link. For commercial queries ("best X," "who should I hire"), this is often the highest-value placement: the model is functionally making a recommendation.

Note the asymmetry between the second and third layers. A citation link earns a possible verification click from the minority of users who click at all; a named mention earns brand placement in front of every user who reads the answer — which, on Overview-covered queries, is most of them. The practical goal of GEO work is both, but the mention is the prize the old rank-tracking mindset tends to miss entirely.

A worked example of the audit

Take a business with five core services. The audit grid is services × questions × engines: for each service, three to five real customer questions, each run through Google's AI surface plus ChatGPT, Gemini, Perplexity, Claude, and Grok, recording per cell — answer present? business cited? business named? competitor named? Even a modest 5×4×6 grid produces 120 observations, which is enough to see patterns: perhaps you are named for one service and invisible for four, or present in Perplexity but absent everywhere else. Those patterns, not any single result, are the honest picture of your Overview exposure — and re-running the grid monthly turns anecdote into trend. This grid is precisely what ClickRadius automates, but the method is valid done by hand; what matters is that it gets done, because the businesses that never measure never notice whose names the answers are teaching their customers.

The window is the story

The most actionable fact in the 48% statistic is timing. Overview coverage tripled in months; user trust in synthesized answers is climbing; and yet most businesses have not adapted — industry estimates suggest the large majority of brands have no AI-search presence at all. The entities that become the default citations for their topics during this formation period will be the incumbents everyone else has to displace later. Displacement is always more expensive than early establishment. That was true of rankings for twenty years; it is proving true of citations now, on a compressed clock.

Frequently asked questions

What percentage of Google searches show AI Overviews in 2026?

Industry tracking data indicates AI Overviews now appear on approximately 48% of Google queries, up from roughly 15% in early 2026 — a more than threefold expansion within months, accelerating sharply around Google I/O 2026 when AI Mode became the default experience.

Do AI Overviews reduce website traffic?

Yes, for most informational queries. Industry estimates put overall zero-click searches near 60% (up from about 45%), and click-through rate for the #1 organic result has fallen from roughly 27% to roughly 11%. However, sources cited inside an Overview gain high-trust exposure, so the strategic goal shifts from ranking under the answer to being cited within it.

How do I get my site cited in AI Overviews?

Make content extractable and verifiable: direct answers under question-style headings, schema markup, and the three signals Princeton's GEO research (KDD 2024) found raise citation likelihood — statistics, attributed quotations, and source citations — plus off-site entity authority, which industry data suggests drives the majority of AI-citation outcomes.

Is your content in the cited class or the buried class? Get your free AI Readiness Score to see how citable your site is across six categories, or review ClickRadius plans for continuous citation monitoring across five live AI engines.