Two of the biggest names in retail made opposite bets on the same idea in the same month.
Shopify launched “Agentic Storefronts”. Merchants can now sell directly inside ChatGPT, Google AI Mode, Microsoft Copilot, and Gemini. No redirect, no website visit, the entire shopping experience happens inside the AI chat.
The same week, Walmart revealed that its ChatGPT checkout experiment converted at one-third the rate of their regular website. Their EVP called the experience “unsatisfying”. They are pivoting away.
One company is going all in. The other just proved it does not work. And I think both are missing the real issue.
Shopping is not a transaction, it is an experience
The first thing that jumps out is that buying stuff is an experience. You want your own agency. You want to feel like you are the one deciding. And most of the time, you want to actually enjoy the process.
Window shopping exists for a reason. People browse not because they need to, but because there is a subconscious pleasure in evaluating, comparing, imagining. You want to feel like you are making the best deal. You want to take your time. You want the colors, the layout, the images, the reviews. That whole dance between you and the storefront is happening at a psychological level that a text box cannot replicate.
When you strip all of that away and reduce shopping to a chat exchange, you suck out the fun part of parting with your money. What you are left with is pure transaction. And nobody enjoys a pure transaction.
It is like food. You go to a restaurant for the presentation, the colors, the smells, the experience of a beautifully crafted dish. You could put the same proteins and carbs in a blender and drink it, your body would not know the difference. But you are not eating for your body. You are eating for your brain. Shopping is the same. The experience IS the product.
But sometimes you just need the jacket
Here is where it gets nuanced. Not every shopping moment is window shopping.
I had this experience recently with Amazon. I needed a winter jacket. Not wanted, needed. There were so many options I was overwhelmed and stressed. I was not there to enjoy the experience, I needed to solve a specific problem. My thinking is very engineering-oriented: I just need quality, features, and price. I did not have time to open twenty tabs and manually compare everything.
So I tried their chat thing. And honestly, it was clunky and stressful. What I ended up doing was using the chat to narrow down the options to five or six, and then I compared them manually myself. A glorified search tab. That is what it was. It did not replace the shopping experience, it just reduced the noise enough for me to do the actual deciding.
So for now, what is available is a glorified search filter with a conversational interface. It makes it slightly easier to narrow down, but it is not actually good at the thing it promises to do.
The shopping profile problem
What nobody in AI commerce is addressing is that the same person shops differently depending on the situation.
If you are browsing for fun on a Sunday afternoon, you want the full experience. The visuals, the comparison, the slow deliberation. A chat interface is the worst possible tool for this because it removes everything that makes browsing enjoyable.
If you are stressed for time and need a gift for tomorrow, your highest priority is speed. You know roughly what you want, you just need the system to find it fast and let you pay. A chat interface could theoretically work here, but only if it already knows your preferences, your budget, your taste. And it does not. Because these systems do not actually know you.
That is the missing piece. A good salesperson in a physical store reads you. They understand what you like, what you dislike, what price range makes you comfortable. There is a whole subconscious dance happening and the salesperson adapts in real time. Current AI shopping tools cannot do any of that because they start from zero every time you interact.
What would actually work
The next evolution is not a chat box that shows you products. It is a system that genuinely understands your preferences, your constraints, and your current shopping context, and then does the hunting for you.
Something that knows you need a jacket, knows you prioritize quality over brand, knows your budget, knows you hate browsing, and comes back with three options that fit along with the reasoning for why each one was selected. Not a search engine with a chat skin. An agent that knows you well enough to shop on your behalf.
We are not there yet. But that is where this is headed, and the companies trying to shoehorn a text box into a shopping experience are solving yesterday’s problem with today’s hype.
The bottom line
As long as humans are the ones shopping and buying stuff and parting with their hard-earned cash, you cannot exclude human psychology from the equation. Trying to replace the shopping experience with a chat interface because chatbots are what everyone is talking about right now is, frankly, kinda stupid. It is shoehorning technology into a use case because the technology is trendy, not because it solves a real problem.
Walmart’s data proved it. 3x worse conversion. Not because the technology failed, but because the approach ignores how humans actually behave when they spend money.
Data beats thesis. And human psychology beats both.
Sources: Search Engine Land, LLM Stats