.jpg)
In late November, my inbox filled with the same headline from a dozen sources: "AI Influences 20% of Holiday Sales." Salesforce had released their Cyber Week report claiming that artificial intelligence drove $67 billion in sales during the five-day shopping marathon
A week later, Etsy’s stock jumped 16% after announcing ChatGPT checkout. Then, Walmart and Target followed with their own AI shopping announcements. Tech Twitter declared that agentic commerce had arrived. You probably came across at least one article which compared this moment to the launch of the iPhone.
And then something strange happened:
Silence.
No transaction numbers, follow-up case studies, or earnings calls saying, “And here’s how much revenue ChatGPT actually drove.” Just press releases, stock bumps, and a lot of flowery, overly optimistic language.
“This is way more hype than reality. OpenAI made several insane, groundbreaking, revolutionary announcements in the last 1 year. How many have truly transformed your life/ecommerce?” A skeptical merchant on Reddit.
It started to feel less like the iPhone launch and more reminiscent of the NFT boom with its overhyped promises and unclear real-world usage.
But are we watching the world’s most expensive PowerPoint presentation, or is there actual value for merchants in AI shopping?
To be clear: this isn't a story about companies lying. It's about something more interesting, and more important for anyone running an online store. It's about understanding the gap between what's technically possible, what's actually happening, and what you should do about it right now.
Let’s start by clearing up what all these numbers actually mean, and how sellers can determine what to do next.
When Salesforce says AI influenced 20% of holiday sales, they’re not saying an AI agent browsed products, compared prices, and checked out on behalf of shoppers like some kind of digital personal assistant from Her.
They’re using a much broader definition.
According to Salesforce's methodology , “AI-influenced” includes purchases where AI was involved anywhere in the journey. That could mean:
All of this counts.
Why? Because it’s actually about AI-researched purchase journeys, NOT agentic commerce.
This lack of definition means it doesn’t actually matter who actually made the buying decision, because it all goes in the same bucket of AI shopping.
This is how hype marketing has always worked; Companies tend to play things up here and there to build hype. But as a merchant, knowing the distinction matters.
There’s a big difference between “AI helped someone decide what to buy” and “AI bought something for them.” Right now, almost all of what we’re seeing is the former.
As one expert noted while analyzing the Shopify, Etsy and ChatGPT partnership: "As exciting as the announcement is for Etsy and Shopify, investors should keep their expectations in check."
Interestingly, following the Etsy + OpenAI launch:
Adobe added fuel to the fire by reporting that traffic from generative AI tools to retail sites grew 670% on Cyber Monday and 760% for the Nov 1-Dec 1 2025 period.
That sounds massive. And technically, it is.
But Adobe also quietly noted something else: “While generative AI traffic remains modest compared to other channels, such as paid search or email, its growth has been notable."
How modest? Adobe's February 2025 report revealed that in July 2024, AI traffic was so minimal it was "too minimal to serve as baseline" for percentage calculations. Their analysts estimated it at roughly 0.1-0.2% of total traffic.
So when AI traffic grew 770% year-over-year, it went from approximately 0.1% to roughly 0.8-1.7% of total retail traffic.
In other words, AI referrals didn’t take over shopping. They went from “barely measurable” to “still super small, but interesting.”
To put that in perspective:
AI traffic, even after growing at DC Flash speed, is still contributing at a microscopic level compared to channels merchants rely on every day.
This doesn’t mean AI shopping won’t matter. It just means we’re a little ahead of the opening scene, but not anywhere near the final battle. We’re just yet to go complete the training montage.
Caila Schwartz, Director of Consumer Insights at Salesforce, stated: "Our data shows that this AI-referred traffic converts at an impressive 8x the rate of traffic coming from social media platforms, proving that AI is quickly becoming the ultimate purchase accelerator."
That sounds incredible until you add context.
Social media is the lowest-converting major channel in ecommerce, sitting at 0.8 - 1.2%. If AI-referred traffic converts at 8x that rate, we're talking about 6.4-9.6% conversion.
Essentially, comparing anything to social is like saying your second-hand car is faster than a bicycle and calling it a breakthrough in transportation.
When you compare AI traffic to other channels, the picture changes:
So it’s decent, but not up there either. The question is: is AI shopping worth the effort that merchants would have to put in?
Yet again, Adobe's own data tells a different story from the narrative that’s been building up. In July 2025, their analysis found that "traffic from generative AI sources was 23% less likely to convert" than non-AI traffic, though that gap was closing rapidly.
Wait, what?
How do we reconcile Adobe showing AI traffic converting worse with Salesforce showing it converting 8x better?
Confusingly enough, both can be true because they're measuring different things.
Adobe tracks all AI traffic, including people just looking products up. Salesforce measures AI-referred traffic that actually clicks through to retailer sites, making them a pre-filtered population that's already expressed purchase intent by leaving the AI platform and visiting the store.
Salesforce is measuring filtered, high-intent clicks. Adobe is measuring everything. Neither means AI is suddenly better at shopping than humans.
Someone who clicks from ChatGPT to a merchant’s site has already done a lot of thinking. They’ve asked questions, compared options, and narrowed choices. That’s high intent behavior.
Another interesting way we can confirm this: Adobe's analysis found that 86% of AI-referred traffic comes from desktop devices, compared to just 34% for overall ecommerce.
Why? Because AI shopping is conversational and research heavy. People are having long back-and-forth discussions with ChatGPT, comparing options, and asking for follow-ups. This is the opposite of impulse buying.
Meanwhile, 70% of actual Cyber Week orders happened on mobile devices. That platform gap tells you these are fundamentally different behaviors: research versus transaction.
All this just means that people who do their homework tend to buy.
If AI shopping was already driving meaningful revenue, you’d expect companies to say so loudly, repeatedly, across company statements and actual discussions.
Instead, we get convenient silence.
In October 2025, Etsy announced ChatGPT Instant Checkout, Walmart teased AI-powered shopping, and Target even rolled out a beta gift finder in the following weeks. We’re midway through January, and we have no solid numbers.
When companies launch something that works, they don’t hide it. They put it in the investor deck, mention it every quarter, and turn the data into multiple case studies.
The absence of follow-up tells you a lot.
When OpenAI launched Shopping Research in late November, Fortune's testing revealed a crucial limitation: "OpenAI warned that Shopping Research can still make mistakes, and you can't buy directly through ChatGPT. The new interface is not connected to the company's Instant Checkout."
For the others, only a small number of products actually supported AI checkout, many features were gated, limited, or not live, and Amazon was busy fighting Perplexity to the death to launch its own Buy for Me program.
This wasn’t Jarvis running your errands. It was more like Clippy with a credit card.
Here’s the simplest way to understand the moment we’re in: AI is becoming a research layer, not a transaction layer.
People use ChatGPT the way they used Google to understand options, compare specs, to narrow down choices, and to ask questions they’d feel silly asking anyone out loud.
To understand how far we still have to go, imagine actual agentic commerce:
You tell your AI: "I need running shoes for marathon training, budget $150, size 10.5, prefer brands that last."
The agent:
All this without you visiting a single website.
What we have now:
Steps 2-6 are exactly what you'd do after a Google search. The AI hasn't replaced the shopping process; It's replaced the search bar. That's useful, but it's not revolutionary.
Buyer behavior shows up clearly in Adobe’s data:
If Google Search was the library, AI is the SparkNotes page summarizing the book and explaining every plot point. You still have to connect the dots and write the book report to submit it.
.jpg)
This distinction matters more than any headline.
Most merchants who’ve been in the game for a while have been using AI. It’s just boring, but extremely effective. You can see this in the:
This is where AI quietly makes money. Most don’t get catchy TechCrunch headlines, but it works.
Shoppers using AI is newer, noisier, and mostly about research. They’re using it for:
That’s valuable, but not important enough at the moment.
The real problem is that when Salesforce reports that merchants using AI tools grew faster, that’s not agentic shopping. That’s just retailers opting for better infrastructure and better tools.
Confusing these two leads to bad, bad decisions.
The claims aren’t dangerous because they’re wrong. They’re dangerous because they pull a lot of attention.
As one merchant platform noted, "Shopify said the update responds to a clear shift away from search pages and toward conversational interfaces where consumers describe intent directly to AI systems.
Is that shift in the room with us?
At under 2% of traffic, "clear shift" is premature, to say the least.
And yet, merchants are asked to reorganize their entire set-up around it. Make content around AI answers. Rebuild entire catalogs. Chase new integrations that dropped last Tuesday.
And why wouldn’t you? Conversion rates already feel hopeless, so you feel like you need to optimize every little thing.
The thing is, the channels that actually drive revenue still demand attention, as compared to the infant that is AI shopping.
Meta still drove 68% of BFCM ad spend. Google still captures high-intent demand. Email still converts at 15-20%. Checkout experience still matters more than discovery tools.
For large retailers with resources, experimentation is fine. They’ve got multiple decades worth of work and money ahead of smaller companies.
But merchants working with tight margins need to make smart bets. Chasing a channel representing 2% of traffic while neglecting channels representing 15-30% is dangerous.
Ignoring solid channels to chase a shiny new channel is how you end up like Blockbuster waiting on late fees while Netflix ships DVDs.
The smart move is boring (which is usually a good sign).
Do these and you should be good.
AI shopping is on its way to becoming bigger. Slowly. Unevenly. And probably not in the way the demos suggest.
Right now, AI is better at helping people decide than helping them buy. It’s upstream of checkout, not replacing it. That doesn’t make it unimportant; It just makes it early.
The mistake isn’t believing AI will matter, it’s believing it already does at scale and making your business revolve around it.
If someone tells you AI drove a ton of holiday sales, ask what they mean by “drove.” Then ask how much actually went through checkout.
For now, the AI shopping revolution can wait. Your fundamentals can’t.
