Across 41 major U.S. platforms with meaningful search behavior, Google still accounts for 73.7% of desktop search activity (Q4 2025). All AI tools combined? About 3.2%. That’s the headline number from new analysis based on Datos’ 2025 desktop panel data and a curated set of 41 domains selected for significant search activity (Source: [5][6]).
And yet, the part that should make a B2B CMO sit up isn’t the 73.7%. It’s what those searches increasingly don’t do. Google Search has become heavily “zero-click” (58% of searches ending without clicks, also cited as 60% overall), and AI search sessions are even more extreme: 93% end without a website visit (Sources: [3][4][1]). The query volume can rise while website traffic stalls. Both can be true. That’s the contradiction—and it’s the story.
Because if discovery is expanding while clicks are shrinking, then classic SEO dashboards are going to keep telling a comforting lie: “visibility down” when brand influence might be flat or up, and “rankings up” when demand capture hasn’t moved at all.
Nut graf: This matters now because early-2026 reporting shows total search usage (traditional engines plus AI/LLMs) is growing—up 26% worldwide and 16% in the U.S. (Source: [1]). Search isn’t dying; it’s spreading out. The practical consequence for demand gen leaders is immediate: teams need a measurement model that treats search as a behavior that happens across engines, commerce sites, social platforms, and AI answer tools—then accounts for the fact that many of those experiences don’t send a click.
The 41-site finding that quietly rewrites “market share”
The new research (attributed in the source material to Rand Fishkin, published March 2, 2026) starts by widening the frame. Instead of treating “search” as only Google versus Bing, it looks at search activity across a basket of 41 domains drawn from the top 250 sites where users commonly query: traditional search, e-commerce, social, reference, travel, real estate, classifieds, and AI tools. The underlying dataset is Datos’ 2025 desktop panel (U.S., EU member countries, and the U.K.). Desktop only. That limitation matters, and it’s stated plainly in the research.
Once that broader definition is used, the distribution looks different. In the U.S. desktop data for Q4 2025, traditional search engines still represent about 80% of searches, but commerce sites make up roughly 10%, social networks about 5.5%, and AI tools 3.2% (Source: [5][6]).
Here’s the pattern interrupt: the “AI is taking over search” story can be simultaneously overhyped and strategically urgent. Overhyped because 3.2% is not a takeover. Urgent because the downstream behavior—no-click answers, fewer visits, more closed-loop discovery—changes what demand gen can reliably measure.
“Search growth” doesn’t mean “traffic growth” anymore
Start with the hard behavioral numbers. Google searches are increasingly ending without a click: 58% (also cited as 60%) are zero-click (Sources: [3][4]). That includes users getting answers directly in the results, refining queries, or choosing features that don’t require leaving Google.
AI tools push that to an extreme. Across AI search sessions, 93% end without a website visit (Sources: [1][4]). Not “lower click-through.” Not “reduced referral.” No visit at all.
So what happens to the standard demand gen equation—rank → click → session → conversion—when the middle of the funnel disappears? It doesn’t just weaken. It breaks. And the replacement metric can’t be “more content,” because experts cited in the brief argue that generic AI content is being deprioritized; what performs better tends to be first-hand experience, opinionated analysis, data-rich insights, and multimedia (Sources: [1][6]).
But the context is more complex. AI referrals do exist. They’re just small: AI referrals are tracked at 1.08% of all website traffic and growing around 1% monthly, with ChatGPT representing 87.4% of AI referral share (Source: [4]). That’s not nothing. It’s also not a reason to gut classic SEO budgets.
AI Overviews, multimodal search, and the new fight: being cited
The most visible driver of “visibility without clicks” on Google is AI Overviews. Their reported prevalence varies by source: Conductor found AI Overviews appearing in 25.11% of searches based on 21.9 million queries, while other reporting cited in the brief puts it closer to ~50% in the U.S. and nearly 55% overall (Sources: [3][4]). The variance is a warning label. CMOs should validate impact in their own category before rewriting roadmaps.
Still, the direction is clear: Google is investing in answer-first experiences at scale. AI Overviews are reported as reaching 1.5–2 billion monthly users and being available across 200 countries and 40 languages (Sources: [2][3][4]). When an interface is that widely deployed, it changes user habits even if marketers don’t like the outcome.
Then there’s multimodal search, which rarely shows up in B2B pipeline reviews but keeps growing. Google Lens processes over 12 billion visual searches monthly, and Circle to Search queries tripled in the past year (Source: [3]). That’s “search happens everywhere” in the most literal sense: the camera is a query box now.
Seen from the other side, this is why the competitive goal is shifting from ranking in ten blue links to being included or cited inside AI-generated answers—often described as GEO/AEO (Sources: [1][2][3]). The question isn’t only “Will this page rank?” It’s “Will the model consult it?” Different game. Different incentives.
A practical executive read: keep SEO, rebuild the scoreboard
For a demand gen leader like David Solis—accountable for revenue, allergic to vanity metrics—the right stance in 2026 is neither panic nor denial. The 41-site dataset says Google is still the main arena on desktop (73.7% share in Q4 2025), and AI tools are still a small slice (~3.2%) (Source: [5][6]). Abandoning classic SEO would be premature.
At the same time, the click-based scoreboard is becoming less honest. When 58–60% of Google searches and 93% of AI search sessions end without a click, influence and traffic decouple (Sources: [3][4][1]). That forces a shift toward measurement that can survive zero-click reality: tracking visibility features, brand/entity mentions, and inclusion/citation in AI answers alongside rankings and sessions (Sources: [2][6][3]).
One more thread from the research points to an operational decision that’s easy to mishandle: governance. Businesses now have to decide how to manage LLM crawlers via robots.txt and the emerging llms.txt standard (Source: [5]). That’s not a developer footnote. It’s a distribution policy. Block the wrong crawler and future “share of model” visibility may never materialize; allow everything and IP exposure risks rise. The right answer depends on category, content sensitivity, and how much the company can win by being cited.
The research opens with a broader definition of search and ends in a more uncomfortable place: the web is still where revenue happens, but it’s no longer where discovery reliably lands. Google’s 73.7% share is real. So is the no-click future. The teams that will look smartest in 2026 won’t be the ones chasing every new interface—they’ll be the ones who accept what the data says, then rebuild their measurement and content strategy to match how people actually search now.
