Google AI Overviews Now Appear in 39% of U.S. Desktop Searches

The part most SEO teams will miss: source ≠ citation

Buried inside the report is a finding that quietly changes how AI visibility should be measured. Comscore found that a site can help inform an AI answer without receiving a visible citation — and for some categories, the gap is enormous.

In lodging-related AI responses measured from December 2025 through May 2026, Tripadvisor appeared as a source link in 61% of responses but as a cited link in only 21%. Hotels.com had 50% source presence but 18% citation presence. Expedia went from 46% to 16%, and Booking.com from 33% to 9%.

This is the difference between being used and being credited. A model can read a page, extract its facts, and synthesize an answer that never names or links the origin. The user gets the information; the publisher gets no attribution and no click. Most AI-visibility tools count the visible citation — which means a large share of real AI influence is invisible to the dashboards SEO teams rely on.

Where AI citations are actually going

When AI answers do cite, the winners are platforms, not publishers. Comscore measured which domains appeared as citations in Google and Bing AI-generated results in June, and the pattern is consistent: YouTube and Reddit are the engines of AI attribution.

YouTube was the most frequently cited domain for entertainment, retail, and financial services searches. Wikipedia ranked first for sports. Reddit ranked near the top across all four categories.

The practical read: if your AI-search strategy is built entirely on getting your own domain cited, you are competing against platforms with massive, frequently-updated, highly-engaged content. A more durable path is to be present on those platforms — and to structure your own content so a model can extract and use it, even when it links elsewhere.

ChatGPT's lead is shrinking — the multi-engine reality

The report also documents a quiet reshuffle in which AI assistant people actually use. ChatGPT still accounted for about half of AI prompt volume in June — but that is down from 70% in January.

Over the same period, Gemini increased from 17% to 30% of prompt volume, and Claude grew from 2% to 11%. Copilot held steady at about 5%. AI assistant reach continued to grow overall, reaching 35% of U.S. desktop users and 29% of mobile users — up from 25% and 12% in January 2025.

For SEO and GEO teams, the implication is straightforward: an AI-search strategy that optimizes for a single assistant is already dated. Visibility now spans Google AI Overviews, AI Mode, Gemini, ChatGPT, Claude, Perplexity, and Copilot — each with its own retrieval and citation behavior.

What this means for SEO

Three conclusions follow from the data. First, search visibility is now a two-part question — where you rank and whether you appear inside an AI answer. Second, “appearing in the answer” is itself two different things: being used as a source and receiving a visible citation. Third, the engines doing the citing are consolidating around a handful of platform domains.

None of this means classic SEO is dead. Google's own guidance on AI Overviews still points to the same fundamentals: create helpful, people-first content that follows the usual ranking systems. The change is that winning a ranking is no longer the end of the visibility question — it is the entry point to a second, less transparent distribution system.

How to measure AI visibility correctly

The source-vs-citation gap means your AI-search dashboard is probably undercounting. Here is how to build a measurement setup that reflects what is actually happening.

1. 1. Baseline Google's own AI reporting. The generative AI performance report in Search Console shows AI Overviews and AI Mode impressions by page, country, and device. Export it before you change anything, and note its limits. Our guide to Search Console AI reports walks through what is included.
2. 2. Track visible citations on representative queries. Use the AI Overview Analyzer across the questions that matter to your audience. Record the answer, the cited URLs, the date, and whether your brand is named — not just linked.
3. 3. Watch for source presence you can't see. Compare what a model says against your content even when it does not cite you. Distinctive, first-party facts and numbers that keep reappearing are your signal that you are being used as a source without attribution — the ghost-citation problem we covered in Ghost Citations.
4. 4. Keep the columns separate. Rankings, AI impressions, citations, source presence, referral traffic, and revenue are related but not interchangeable. A single blended “AI visibility score” hides more than it reveals.
5. 5. Track rising queries, not just rankings. As AI coverage expands, new question-style queries appear. Use a keyword search volume checker to spot demand before it shows up in your rank tracker.

What to do now

The strategy that wins in an answer-first, multi-engine, source-vs-citation world is not dramatically different from good SEO — but the priorities shift. Here is where to put the effort.

What to watch next

Comscore's report is a snapshot through June 2026, which means the next quarterly release will be the real test of the trend. Watch whether AI Overview coverage keeps climbing, whether Bing's Copilot share accelerates, and — most important — whether the source-vs-citation gap narrows or widens.

Two developments would change the picture quickly. If Google starts reporting AI click-through data alongside impressions, the coverage-vs-outcome confusion disappears. And if more publishers adopt structures that make source attribution explicit — or if engines begin citing sources more consistently — the gap between being used and being credited could close.

Until then, treat AI visibility as a three-part measurement — coverage, source presence, and citation — and build content that earns the last of those without depending on it.

Frequently asked questions