“People who shop with AI? Still a fringe thing, surely.” That was me, not long ago.
Lately the headlines nag at it — “AI referral traffic up 4x”, “shoppers from AI buy more” — and it’s easy to wave them off. I did. I figured a few early adopters ask ChatGPT for tips, sure, but almost nobody follows through to an actual purchase.
So more AI traffic just meant more browsers, more tire-kickers. Harmless, ignorable.
But then I looked at an Adobe dataset built on over a trillion visits, and the “fringe” story stopped holding up.
Because the shoppers AI sends didn’t just grow in number. Somewhere in the last year they quietly turned into the best customers on the site.
First The Volume: AI Traffic Roughly Quadrupled In A Year
The study is Adobe Analytics’, published in April 2026, based on more than a trillion visits to US retail sites (the study). That’s a big enough sample that “rounding error” stops being an easy dismissal.
And the growth is steep. Traffic from generative AI was up 393% year-over-year in Q1 2026 — call it roughly 4x in twelve months. Over the 2025 holiday stretch it ran even hotter, up 693% on the year before.
Sure, a big percentage off a tiny base is nothing. But 4x off a trillion-visit base isn’t a rounding story anymore. This has left “a handful of early adopters” behind.
In One Year, The Ranking Flipped
Growth is meaningless if none of them buy, though. That’s the real question — and it’s where this gets interesting.
Conversion rate is just the share of visitors who actually buy something. As of March 2026, Adobe’s data has AI-referred shoppers converting 42% higher than typical visitors. Revenue per visit — the average money each visit brings in — ran 37% higher too.
Here’s the part that got me: a year earlier, it was the exact opposite. AI-referred visitors converted 38% worse, and on revenue per visit, regular visitors beat them by 128%. So the crowd that used to be “shows up, never buys” flipped, inside a year, into the highest-value traffic on the page. If you’re still filing AI shoppers under “not serious”, that’s a rating you forgot to update.
Why So High-Intent? They Do The Deciding Before They Arrive
So why does AI traffic buy like this? The behavior data makes it click.
- Time on site runs 48% longer than a normal visit
- Pages per visit are 13% higher
- Engagement rate — the share who actively interact rather than bounce — is 12% higher
The shortlisting, the comparing, the “but does it do X” doubts — a lot of that happens inside the chat, before the click. So the person who lands on your page has basically finished shopping and walked in to buy. Less hesitation, higher intent, more “just take my money.”
Consumer surveys point the same way: 39% of people already use AI for online shopping, and 85% say it made the experience better. That’s not a fad spiking — that’s a buying habit resettling.
But For A Lot Of Stores, The Front Door Won’t Open
Which would be a tidy “AI traffic is great, the end” story. Except the same Adobe study has a number that makes you wince.
These high-intent buyers keep arriving, and about 34% of product pages can’t be reached by AI systems at all. Another 25% or so of homepages and category pages aren’t set up for AI either.
just means the AI can actually read what’s on the page. A human sees it fine, but if the key details live inside an image, or get painted in later by JavaScript, the AI often can’t pull them — and that’s a wall.
Think of it as leasing prime real estate and then wiring the automatic door to stay shut. Your highest-spending customer walks right up, and one time in three the door doesn’t open, so they leave. Great traffic that can’t be read turns straight back into lost sales.
One honest caveat: Adobe sells site-optimization tools, so “make your site AI-readable” happens to be exactly what its product does. Read the conclusion knowing that. Still, a trillion-visit sample and a full reversal in traffic quality are hard to explain with sales talk alone.
Conclusion: Before You Chase Rank, Check That AI Can See You
So what does this actually change? Mostly the order of operations. AI traffic is already here, and it’s already your best-converting traffic. But the door those buyers walk up to might be shut from the AI’s side. The first move isn’t the trendy keyword or the search rank — it’s the dull check of whether your main pages can even be read by AI.
I came in assuming AI shopping was a next-year problem, and the numbers talked me out of it. Two things are worth doing, and only two:
- Have someone confirm your top product and category pages are readable to AI — key details as real text, not buried in images or script-rendered after load
- Then watch how the AI actually names and routes you, because “made it readable” and “actually getting picked up” are different claims
The high-intent shoppers are already showing up. The only question is whether your door opens when they do. Maybe I’m wrong and your category is the exception — but “check it” beats “assume it”, and checking is cheap.
Sources
- Adobe Analytics (2026), “AI traffic surge: retail sites still aren’t machine-readable”, April 2026, adobe.com (over 1 trillion visits to US retail sites. Generative AI traffic up 393% year-over-year in Q1 2026, and up 693% over the 2025 holiday season. As of March 2026, AI-referred shoppers convert 42% higher and show 37% higher revenue per visit; one year earlier they converted 38% lower and regular visitors led revenue per visit by 128%. AI-referred visits run 48% longer, 13% more pages, 12% higher engagement. About 34% of product pages are not accessible to AI systems and roughly 25% of homepages/category pages aren’t optimized for AI. Caveat: Adobe sells site-optimization tools.)
- Adobe (2025), “Adobe Analytics: Traffic to U.S. Retail Websites from Generative AI Sources Jumps 1,200 Percent”, March 17 2025, blog.adobe.com (consumer survey: 39% already use AI for online shopping; 85% say it improved the experience.)
