Everyone keeps repeating the same story about AI search: traffic is surging, organic is collapsing, the ground is moving under your feet. It's a clean narrative. It's also mostly wrong, and reporting it to your boss as fact is how you end up looking foolish in a quarter.
A fresh dataset says the panic is misplaced, but it also exposes a measurement mistake that almost every marketer is making right now. Fix it this week and you'll be reading your AI search numbers more accurately than most agencies charging you for the privilege.
SEO researcher Aleyda Solis pulled April 2026 Semrush data for 40 top U.S. sites across four verticals (Travel, Finance, Real Estate, and Shopping/Retail) and compared two things almost nobody looks at side by side: where AI actually sends clicks, versus which pages AI systems cite in their answers.
Start with the headline. Across all 40 sites, organic search drove roughly 792 million visits. AI traffic drove about 7.4 million. That's organic outweighing AI referrals by something like 108 to one. If anyone in your leadership meeting is treating AI referrals as a replacement for SEO today, the data isn't on their side.
Here's the part nobody's talking about. The pages AI sends clicks to and the pages AI cites in answers are almost two different worlds. Homepages and brand-entry pages got 57.7% of AI traffic but only 3% of AI citations. Discovery and evaluation pages (guides, category pages, comparison content, support docs) got just 8.9% of AI traffic but a massive 57% of AI citations.
The pages doing the heavy lifting to get your brand into AI answers are barely showing up in your traffic reports. They're invisible to anyone who only watches referral clicks. The comparing and narrowing happens inside the answer. By the time someone clicks, they've already decided, so the click lands on a homepage or buy-now page. The work that earned the recommendation is upstream.
The clearest proof is Stripe. It accounted for about 43% of all Finance AI traffic, but almost all of those visits hit login pages, OAuth redirects, and checkout tokens. Meanwhile, Stripe's cited pages were almost entirely small-business education content. One spreadsheet, two completely separate stories. A spike in AI traffic to a login page is not a discovery win.
Don't report one blended AI search metric. Split it into three layers. AI Presence: does your brand show up in relevant answers at all. AI Citations: which specific URLs AI systems reference. AI Traffic: which URLs actually receive clicks. Then add one column to whatever you track: page type. Tag every URL as homepage, category, product, guide, support, checkout, or auth/redirect.
Add one more column worth its weight: traffic owner. In retail, travel, and marketplaces, AI can recommend your brand and route the click to Amazon, Walmart, or an OTA. Volume alone won't tell you who captured the customer. You need to know when AI is creating demand you don't get to monetize.
If the pages that get cited aren't the pages that get clicked, your optimization has to split too. Strengthen the pages AI cites: guides, comparison pages, category and listing pages, support and policy content. Then make sure the pages that do get the click are ready to convert and that you own the traffic, not a marketplace.
Open your AI search report today. If it shows a single AI traffic number, you're flying blind. Add a page-type column, separate citations from clicks, and rerun the story. You'll almost certainly find that your most valuable AI work is happening on pages your dashboard never flagged.
"Organic search is the scale channel. AI citations are the visibility layer. AI traffic is the endpoint. The marketers who win the next 18 months stop confusing the three."