Datos Caught AI Search at 1.72% of Visits While Google Climbed Back to 94.3%
Datos, a Semrush company, published its Q1 2026 State of Search report this week, drawing on clickstream data from millions of desktop users across the US, EU, and UK. AI tools hit 1.72% of desktop visits in the EU and UK, up from 0.98% a year earlier. In the US, the figure rose from 1.31% to 1.65%. Google's organic share climbed back to 94.3% in the US after a year of softening.
The headline number is the one most newsletters will run with: AI is still under 2%. The growth rate is the number worth budgeting against.
The growth curve, not the floor, is the number worth budgeting against
Monthly data from the same Datos panel shows AI tool share at 0.54% in January 2025. By March 2026, it sat at 1.08%. That is a doubling in fourteen months. If you draw a straight line through that curve (and the curve has stayed close to linear, not exponential, for three quarters running) AI tools land somewhere in the 2.5% to 3.5% range by Q4 2026.
Rand Fishkin, who co-published the report with Datos, called the under-2% figure "shocking" given the cultural attention on AI. He's right that the floor is lower than the hype suggests. But the YoY change in EU/UK was 75%, according to PPC Land's coverage of the report. A 75% YoY growth rate, sustained for two more quarters, doubles the share again by next April.
That's the question marketing leaders should be modeling. Not "what is AI search today?" but "what does my organic acquisition curve look like if AI sessions hit 4% of visits by next year, and 8% by 2028?"
The Datos report also broke out Google's own AI Mode separately. It sits at 0.16% in the US and 0.21% in the EU/UK. Tiny in absolute terms, but Google's AI surface is being measured inside the AI tools bucket and on top of Google's existing 94.3% organic share. If AI Mode adoption picks up the way Gemini's standalone product has, the conversation about "AI cannibalizing search" gets a lot more complicated, because most of that cannibalization stays inside Google's own walls.
In most cases I've seen, the answer is: not catastrophic for SEO, but real enough that you should be allocating budget to AI visibility now, while it's cheap.
Google's 94.3% recovery is a tailwind, not a guarantee
The other line buried in the Datos data is that zero-click searches dropped. In the US, the share of searches ending without a click fell from 24.5% in December to 22.4% in March. In the EU and UK, the drop was sharper: from 22.5% to 19.6%. Organic click share in the US rose from 42.0% to 44.9% over the same window.
That's a real recovery for publishers and SEO programs that survived the AI Overviews squeeze in 2024 and early 2025. Semrush has been tracking the same broad pattern through its own clickstream data. The combination of Google's organic share rebounding to 94.3% and zero-click sessions falling has given organic teams roughly six to nine months of breathing room. We've also seen the click-loss debate sharpen elsewhere, including a recent randomized study that pegged AI Overview click loss at 38% for queries where Overviews actually appeared, which is a useful reminder that aggregate recovery and per-query damage are two different things.
I think most teams will misread this window. The temptation will be to interpret the recovery as a sign that AI Overviews "didn't really hurt that much" and pull back on the GEO and AI visibility work that started in 2025. That's exactly backward. The recovery is the window in which to invest, because the share that's coming back to organic is the share you can capture before AI tools take a bigger cut next year.
There's also a methodological caveat worth saying out loud. The Datos panel is desktop-only. Mobile, where roughly two-thirds of searches happen, isn't in the dataset. The recovery numbers might look different on mobile. Worth pressure-testing against your own Search Console before you celebrate.
Why the conventional reading of the report goes wrong
The way most marketing leaders are going to interpret "AI is under 2%" is: pause the AI search investment, defer the GEO hires, focus on what's working today.
A few problems with that read. First, the share of AI-driven traffic is heavily skewed by vertical. According to a Semrush clickstream study, referral traffic from ChatGPT to websites grew 206% YoY (Jan 2025 vs Jan 2026). For some categories (B2B SaaS, developer tools, certain consumer research queries) AI sources are already in the 4 to 7% range. The 1.72% aggregate hides a long tail of categories where the number is closer to 6%, and a few where it's already above 10%.
Second, AI search traffic is still cheaper to win than organic. A lot of the optimization techniques (clear answer extraction, structured data, source authority) are the same ones that help featured snippets. You can do both at once. Splitting the work into two roadmaps is mostly wasted effort.
Third, the panel mix is changing fast. Per Semrush's 2026 AI SEO statistics roundup, Gemini's user base has roughly tripled in the last twelve months, while ChatGPT's share of the AI category has fallen from above 87% to closer to 68%. If your AI search work was built around ChatGPT alone, you've already missed Gemini and Claude. The Datos number doesn't tell you that. The market share data does.
And honestly, search isn't only happening in those tools either. SparkToro's research on where search actually happens across 41 websites shows that a meaningful chunk of category-specific intent now starts on Reddit, Amazon, YouTube, and a handful of niche communities. Those are the sources AI tools are training on and citing. If you're not present there, you're invisible to the AI layer too.
Where to actually move budget over the next two quarters
If you're running an SEO program in 2026, here's the resource allocation I'd argue for given Datos' Q1 numbers.
Keep core SEO investment flat or slightly increased. The 94.3% Google share and the organic click rebound mean every dollar you put into Google organic is currently earning more than it did in mid-2025. Don't use the AI hype as cover to cut the budget that's actually paying.
Allocate roughly 10 to 15% of your SEO team's hours to AI visibility work. That includes Schema.org markup tightening, FAQ structuring, source citation patterns, and direct presence on the platforms where your category's questions get asked. Reddit, Stack Overflow, niche forums, the long tail of community sites SparkToro mapped. The platforms feeding AI tools are not just Google.
Set a benchmark for AI session share. By Q4 2026, if your category isn't pulling at least 3% of organic-equivalent sessions from AI tools, you're behind the Datos curve. By Q2 2027, the benchmark to clear is closer to 5 to 7%, depending on category.
And audit your conversion attribution. From what I've seen, when a session lands from ChatGPT or Perplexity it currently shows up as direct traffic in most analytics setups roughly 50 to 60% of the time. You can't optimize what you can't see. Until your attribution can credit the AI source, every report you build will quietly understate the channel.
The under-2% number isn't the story. The doubling rate is. Plan capacity against the trajectory, not the floor, or you'll be the team scrambling in eighteen months when the floor isn't 2% anymore.
