4700%: E-commerce Traffic Is Carving a New Riverbed

A few days ago, a friend of mine who runs an online store joined me for tea. He said something that stopped me cold: "I don't think I need to decorate my storefront anymore."

Why not?

Because, he said, more and more customers never come in through the search engine door. They ask an AI first — "help me pick a sunscreen for sensitive skin" — and the AI just hands them an answer. However beautiful your product detail pages are, those customers never see them.

I figured this deserved a proper run through the numbers. So I dug through a stack of reports — from McKinsey, NVIDIA, Adobe, Gorgias. When I finished, I was more convinced than ever:

E-commerce traffic is carving a new riverbed. The water is the same water; the channel is no longer the old one.

Start with the Big Number

What does "a new riverbed" mean? Let me give you three numbers first.

In 2024, the global e-commerce AI market was worth roughly $7.25 billion. Precedence Research projects that by 2034, that figure will grow to somewhere between $64 billion and $75 billion, a compound annual growth rate of 23.6%.

A decade. Tenfold.

And that's just the money spent on tools. McKinsey ran another calculation: by 2030, retail transactions brokered or influenced by AI agents could reach $1 trillion in the US market alone, and $3 trillion to $5 trillion globally.

What does $3 trillion mean? Roughly the GDP of several countries, stuffed into a single drawer labeled "AI does your shopping for you."

So stop asking "should we use AI?" NVIDIA's early-2025 survey found that more than 80% of retail and consumer-goods companies are already deploying or piloting generative AI, and nine in ten plan to keep pouring money in.

At this point, that question has answered itself.

The Real Divide Is Somewhere Else

But every trend has a flip side.

McKinsey's State of AI 2025 report contains a stinging comparison: among large companies with over $5 billion in annual revenue, nearly half have already scaled AI across the business for real. And small companies under $100 million in revenue? Less than a third.

Everyone is using it. Big companies have turned it into systems; small companies have turned it into toys.

Where's the gap? Having read the data, I'd say it comes down to two things.

The first is people. In NVIDIA's report, "lack of AI talent" jumped from 31% last year to 46%, becoming the number-one blocker in just twelve months. Gorgias surveyed 400 e-commerce decision-makers, and 71% of brands said they plan to hire specifically for AI-related roles in the next 12 months.

The second is data. McKinsey's respondents said 40% of their time goes to low-value work like consolidating data and connecting systems. Think about it: the grunt work that AI was supposed to take over is still being done by humans, because the data plumbing underneath was never connected in the first place.

AI doesn't lack enthusiasm. What it lacks is a foundation — and the foundation is people and data.

Where Does the Money Actually Come Back?

Alright, let's talk about something more encouraging. With all this investment, where's the return?

McKinsey's research on personalization offered a range: personalization typically drives a 5% to 15% revenue lift, with the best performers reaching 25%. Across all functions, marketing and sales have the hardest evidence of AI gains — 67% of marketing and sales teams said AI genuinely added money to their bottom line over the past year.

Customer service is even more dramatic. Rep AI analyzed 17 million shopping interactions: customers who chatted with an AI converted at about 12.3%; those who didn't, at 3.1%.

Four times higher.

Over at Gorgias, 79% of brands said AI-powered conversational commerce lifted their sales. In NVIDIA's report, 91% of retail and consumer-goods companies said AI is helping cut supply chain costs, and demand forecasting is the most-used AI application in the supply chain at 64% — nearly double the runner-up.

Do you see the pattern? What AI eats first are the jobs with "clear rules, abundant data, repetitive labor": forecasting inventory, writing ad creative, answering support tickets. In NVIDIA's survey, 67% of digital commerce companies use AI for marketing content creation — ahead of ad targeting, recommendation systems, and support chatbots.

Content, of all things, came first. Imagine that.

The Riverbed Really Has Shifted

Now, back to my friend's line: "the storefront doesn't need decorating."

Adobe has been watching over a trillion visits to US retail websites, and it found something: traffic coming from generative AI grew 4,700% in a year.

A 4,700% growth rate. And these customers are astonishingly high quality — they stay 32% longer than visitors from paid search, email, or social media, browse more pages, and bounce less.

McKinsey's data confirms it: among people who have tried AI search, 44% already treat it as their primary way of searching. Brands cited in Google's AI Overviews get click-through rates 35% higher than ordinary search results — that's what Yotpo found after analyzing 53 million reviews.

So whom does AI cite? Triple Whale ran a ruthless experiment in early 2026: tracking more than 600,000 AI-model citations answering e-commerce questions. Number one: Reddit, at 28.8%. Number two: Alibaba, at 15.7%. Then Wikipedia, Yahoo, YouTube.

Reddit. A forum.

Why? Think about it: when a user asks an AI "is this massage gun worth it," whom does the AI trust most? Places where real humans argue. Reddit is nothing but real human discussion — praise and complaints alike, no advertorial tone.

AI is selecting "human-sounding" information on behalf of humans. So your best ad placement in the future may not be inside an ad system at all, but inside real users' conversations.

Customers Want AI — But Only Half of It

And on the consumer side? We have to look at both ends.

Yotpo's research says 58% of Gen Z already use AI for product discovery while shopping, and interest sits between 52% and 66% across every age group — this is not a young people's monopoly.

But Gorgias also found that 54% of customers, when something goes wrong with an order, still want a human. And 41.5% of e-commerce professionals worry that AI simply can't fully resolve customers' problems.

Attitudes toward reviews are fascinating too. 66% of shoppers hesitate to place an order when a product has fewer than 5 reviews; 70% of Gen Z actually trust products with "both good and bad reviews" more; half of millennials deliberately seek out the middle-of-the-road and negative reviews.

As for AI summarizing reviews for you, the mainstream choice across both generations is strikingly consistent: not AI summaries alone, not raw reviews alone — read both.

Translate the consumer attitude and it says: AI, you help. But the final call is still mine.

Anthropic's global survey of 81,000 people agrees: 67% of AI users feel positive overall, but the number-one worry is "unreliability" — 26% mentioned hallucinations, wrong answers, and the burden of "having to double-check everything myself."

The enthusiasm is real. So is the wariness.

Back to That Cup of Tea

The tea finished, I told my friend: yes, you still need to decorate the storefront.

But the logic of decorating has changed. It used to be decorating for the search engine crawlers — stacking keywords into titles. From now on, it's decorating for the AI, and for the real humans behind the AI, so they're willing to talk about you, cite you, argue over you. Your product descriptions, your user discussions, your replies to bad reviews — all of it is "decoration" in the AI's eyes.

Thirty thousand years ago there was already a river, and the water has never stopped flowing. It was never the water that changed course; it was the terrain.

This time, the terrain has already moved. 4700% is the signal.

May you dig your new channel early.