Today’s wrambling is going to be exactly that, a rambling. We need to talk about generative AI. Because no doubt, like many of you, my brain hurts from trying to make sense of all the noise.
So today I’m going to say something that might surprise you. I think we need to pump the brakes.
I know, I know. That’s not the sexy take everyone wants to hear right now. After all, the headlines coming out of Black Friday 2025 were screaming about AI’s “transformative” impact on shopping. According to Adobe Analytics, AI-driven traffic to retail sites soared 805% compared to last year. AI “agents” influenced $14.2 billion in online sales globally, per Salesforce. ChatGPT just launched its shopping research feature. Walmart began testing ads in its Sparky AI shopping assistant. And all that was in just one week of headlines!
The narrative is clear. AI is the talk of the town. Don’t get me wrong. AI is hands down the most important innovation of the last 30 years. Every retail executive needs to be thinking about it but the degree to which everyone should just rush into it guns ablazing is about as grey as Fletch’s charcoal.
The Numbers Tell a Different Story
Let’s start with that 805% increase in AI-driven traffic. Sounds impressive, right? But when you dig deeper, I think it’s actually surprisingly low.
Here’s why: We’re essentially in the early innings of consumer-facing AI shopping tools. Remember when Walmart launched Sparky and Amazon launched Rufus? That wasn’t that long ago. So we’re looking at growth from a tiny base to what is likely still a relatively small number.
To put this in perspective, I looked back at Buy Now, Pay Later adoption. According to CFPB data, originated BNPL loans grew 970% from 2019 to 2021. AI shopping tools are supposedly the future of commerce, backed by billions in investment from the world’s largest tech companies, and they’re growing slower than a payment method that’s basically just installment plans with better marketing, and one that at the time was also further along its growth trajectory.
That should give us pause . . . at least pause enough to think about what matters and what doesn’t in this conversation. The rate of AI’s impact on commerce is fast but is it as fast as we all want to think that it is? And which side of it, the search impact or the agentic impact, do we need to be concerned about the most?
The $64K Research Question
To answer that, let’s analyze ChatGPT’s new shopping research feature. On the surface, it makes total sense. Instead of sifting through dozens of sites, you describe what you’re looking for, and the AI builds you a personalized buyer’s guide.
But here are the questions I keep coming back to: How many product searches are actually that research-intensive? And at what stage do research searches happen within the purchase journey? And how do we factor in consumer behavior across the various search versus commerce platforms?
For these answers, we have to analyze what were, for all intents and purposes, the two pre-GPT search behemoths: Google and Amazon.
According to SparkToro research conducted in 2024, it can be assumed that whenever someone is buying a product for the first time, there is generally some amount of information seeking being conducted within initial searches. SparkToro estimates that across general Google search behavior:
* About 53% of searches are informational (general research, learning).
* About 32% are navigational (getting to a specific site or brand).
* About 14% are commercial (product/service investigation but not yet “buy now”).
* Less than 1% are transactional (strong “buy / sign up / book” intent)
However, not all shoppers start their products searches on Google. Many start on Amazon, as many as 60%. But, on Amazon, the behavior is different. There, according to multiple sources, the range of search behavior on Amazon can best be categorized as:
* 70–90% “transactional or near-transactional” (including branded and very specific product searches).
* 10–30% “research/consideration” (category exploration, feature comparison, use‑case discovery).
So, what this tells me is that, depending on where one is in his or her buying journey, he or she will either want a comprehensive buying guide or they may simply just want to complete a transaction. The degree to which people want to do both in one all encompassing platform is the unknown, and something which Google has heretofore never been able to unlock, which is also why Amazon has maintained its stranglehold on transactional product search.
People aren’t going to Amazon because they want to research. They’re going to Amazon because they know what they want and they want it fast.
So when I see all this excitement about AI shopping assistants helping people research complex purchases, i.e. finding the quietest cordless vacuums, comparing three different bikes, choosing between electronics with specific specs, I have to ask: How big is that use case really? And more importantly, is it big enough to justify the massive investments and strategic shifts retailers are making to accommodate these platforms?
Because last time I checked, pre-GPT, Amazon has done exponentially more in commerce than Google ever has.
Where Retailers Actually Have Power
Here’s where I get fired up, because I think retailers have way more power in this equation than is being discussed.
When Walmart announced this week that it is testing sponsored prompts in Sparky, I was 100% supportive. And here’s why: Walmart is focusing on creating a great generative AI search experience on its own site.
Think about the logic here. Walmart and Amazon both have massive marketplaces. They’re the everything stores. They have the traffic, the product selection, the fulfillment infrastructure, and now they’re building AI-powered search experiences directly into their own properties.
All of which begs another question: Why do we even need ChatGPT and Google for agentic commerce?
Amazon previously carved out a huge chunk of transactional product search traffic from Google by creating a better, more transactional experience. Why can’t Amazon and Walmart continue to do that in the AI era?
The answer is: They can. And they should.
The Pragmatic Approach: Five Things Retailers Should Do Right Now
So what does a pragmatic approach to generative AI actually look like? Based on everything we’re seeing, here’s what I’d recommend:
First, invest in AI on your own site. We already know generative AI searches convert at higher rates than traditional search. So make your on-site search experience better with AI. Make it conversational. Make it helpful. Make it feel like ChatGPT, but for your products, on your property, where you control the experience and capture the data.
Second, watch the traffic. The line you should be monitoring in your Monday morning meetings is: What traffic and volume is actually coming from ChatGPT, Google’s AI shopping, and similar platforms? My hunch is that it’s pretty small right now. If anyone out there at the retailers wants to send me a private message and prove me wrong, please do. But until I see data suggesting otherwise, I think the traffic from these external AI platforms is minimal. For example, the referred traffic to Omni Talk from ChatGPT this past month was less than 3%. If I factor in organic traffic, that number falls to less than .1%. And we just run a news business. Meaning there shouldn’t be much we need to do to make our content GPT-ready.
Third, think like an ad business. If you do build AI shopping experiences on your own site, monetize them. Walmart’s approach with sponsored prompts is exactly right. You’re converting traffic at higher rates, you’re getting more ad dollars, and you’re keeping customers on your property. That’s a win-win-win.
Fourth, be selective about integration. Watch that traffic flow from the GPT-like platforms, and use that data to decide if and when you actually need to connect with them. Don’t just rush to connect all your inventory, all your systems, all your customer data to every AI platform that comes along because “agentic” is the new buzzword. Be strategic. Be thoughtful. Ask yourself: Which Pandora’s boxes do I actually want to open?
Fifth, remember the first order of operations should be that you’re building an external search amplifier, not a replacement web experience. The best way to think about these AI tools, at least for now, is as search amplifiers. They help customers find what they’re looking for faster and more efficiently. They’re not replacing the fundamental shopping experience. They’re not replacing your brand. They’re not replacing your customer relationships. They’re amplifying search. Which is why Target, in my opinion, made the absolute wrong move by approximating its mobile app experience inside of ChatGPT. Target effectively ceded control of one of its most valuable assets by placing an intermediary between it and the Target customer.
We Have The Power
As retailers, we have more power than we think. Customers aren’t just looking for technology and speed. They’re looking for quality, trust, and value. And those are things on which traditional retailers can absolutely compete.
The AI shopping narrative right now is being driven by Big Tech. They’re the ones setting the agenda, controlling the platforms, and writing the headlines. There’s a real risk in retailers just reacting to every announcement, every new feature, every supposed whiz bang, without stopping to ask the hard strategic questions.
Questions like: What traffic are these platforms actually driving? What’s the conversion rate? What’s my margin? What customer data am I giving up? What control am I ceding? What happens if the platform changes its terms or goes in a different direction?
These aren’t sexy questions. They don’t make for exciting conference presentations or viral LinkedIn posts. But they’re the questions that determine whether your AI strategy actually makes money or just makes noise.
What This Means for Your Business
So what should retail executives actually do with all this?
Well, if you’re a major retailer like a Walmart or a Target, invest heavily in your own AI capabilities. Build the best damn AI shopping experience you can on your own site. Monetize it through ads. Test, iterate, and improve. Watch what traffic comes from external platforms, but don’t rush to integrate with every new AI shopping tool that launches.
If you’re a specialty retailer or a regional player, focus on what makes you different. Use AI to amplify your unique value proposition, whether that’s curation, expertise, service, or community. Don’t try to out-Amazon Amazon or the GPTs. Build AI experiences that reinforce why customers come to you instead of going to the everything stores, both now and in the future.
If you’re an upstart brand, think carefully about distribution. Yes, you want to be discoverable on AI platforms. But you also want to maintain direct relationships with customers. Invest in your own site experience. Build your own AI shopping tools. Don’t let AI platforms become the new Amazon marketplace problem or another intermediary between you and your customers that takes a cut and controls the relationship.
Because if you think about it, if you go in the opposite direction, it effectively means that there really never has been any reason for the industry as a whole not to list its products for sale on Amazon.
Let that last statement sink in.
Slow & Steady Wins The AI Arms Race
Look, I’m not saying AI isn’t going to be important in retail. It absolutely is. I’m not saying retailers should ignore it. They absolutely shouldn’t.
What I am saying is this: Slow down. Be strategic. Be pragmatic. Don’t believe every headline. Don’t chase every shiny penny on the ground waiting to be picked up. Don’t give away your power.
The retailers who win in the AI era won’t be the ones who rush to connect everything to ChatGPT in 2025. They’ll be the ones who asked hard questions, made strategic bets, invested in their own capabilities, and maintained control over their customer relationships and experiences.
So before you sign that next big AI partnership deal, before you redirect your team to work on integrating with the latest AI shopping platform, ask yourself: What problem am I actually solving? What data supports this investment? What’s my alternative? And what happens if this doesn’t work?
Because here’s what I’ve learned over my now almost 30 years of studying and watching this industry: The best retail strategies are usually the boring ones. The ones that focus on fundamentals. The ones that improve the experience for customers on your own properties. The ones that make money.
And right now, based on everything I’m seeing, the pragmatic approach to AI in retail is pretty simple: Build it on your site. Watch the data. Stay flexible. And don’t give away the farm to Big Tech just because everyone else is doing it.
Get full access to Walton's Weekly Wramblings at chriswalton.substack.com/subscribe
