When the Customer Is a Machine
The percentage of Americans seeking AI advice on buying decisions increased from 12 percent to 35 percent in just the past year.
Part 3 of Bill Sparks' series on how AI is becoming the new middleman between brands and customers. This installment looks at the different strategies that are being developed to turn AI shopping advice into sales.
According to Adyen’s 2026 Retail Report, 35 percent of American shoppers now use AI assistants for shopping. A year earlier, that figure was 12 percent. More than half said they would be willing to let AI handle the entire purchase process. Morgan Stanley Research projects that AI shopping tools could drive up to $385 billion in U.S. online sales by 2030, accounting for as much as a fifth of total e-commerce spending.
Those numbers mean something specific for brands. The question is no longer whether customers will use AI to shop. It is what happens to your brand when they do.
The reality check
The early bets on AI-powered checkout have been humbling. In November 2025, OpenAI launched Instant Checkout, letting users buy products from Walmart, Etsy, and Shopify merchants without leaving ChatGPT. The logic was straightforward: if people are already asking ChatGPT what to buy, let them buy it right there. Walmart made roughly 200,000 products available through the feature.
The results were not kind. Walmart’s EVP of product and design, Daniel Danker, called the in-chat checkout experience “unsatisfying.” Checkouts completed inside ChatGPT converted at about one-third the rate of sessions where users clicked through to Walmart.com. Customers were willing to ask an AI what to buy, but when it came time to enter a credit card, they wanted the familiar checkout they already trusted.
OpenAI read the data and shifted course. It began phasing out Instant Checkout in favor of merchant-controlled checkout, handing the transaction back to the retailer’s own site. The lesson was clear enough: consumers will let AI help them choose. They are not yet ready to let it handle payment.
ChatGPT Shopping now processes roughly 50 million shopping queries a day, according to a study by OpenAI’s economic research team and Harvard economist David Deming. That is enormous. But the queries are about discovery, not purchase. The AI is a shopping companion, not a cashier.
How the plumbing works
Two competing protocols are being built to connect AI agents with retailers. The Agentic Commerce Protocol, or ACP, was co-developed by OpenAI and Stripe. It handles checkout sessions and secure payment transmission and has signed more than 25 partners, including Salesforce, Squarespace, and Adobe Commerce. Think of it as the payment API for letting an AI agent complete a transaction on a retailer’s site.
Google and Shopify took a different approach with the Universal Commerce Protocol, or UCP. Where ACP handles the transaction, UCP covers the entire journey from product discovery through post-purchase support. Its March 2026 update added multi-item carts, live catalog queries, and loyalty program integration. Partners include Visa, Mastercard, and more than 20 others.
Shopify is making an infrastructure play that extends beyond its own merchants. At the National Retail Federation’s annual conference in January 2026, Shopify announced its Agentic Plan, which lets brands that do not use Shopify join its product catalog anyway, making them visible to AI shopping agents through Shopify’s infrastructure. The message is strategic: Shopify wants to be the catalog every AI agent reads, regardless of whose platform runs the store.
What changes for brands
Data is all AI sees. Sponsored listings, ads, and keywords are meaningless. Its recommendations are based on specifications, descriptions and structured attributes.
The traffic pattern is already shifting. ChatGPT drives roughly one in five of Walmart’s referral clicks, according to Similarweb data, with similar shares for Etsy and Target. Amazon, which blocked many AI crawlers, saw its ChatGPT referral traffic fall nearly 18 percent over the same period. The brands that get recommended in AI answers are getting a meaningful share of discovery traffic.
But an AI agent does not browse the way a human does. It does not scroll past sponsored listings, notice banner ads, or get pulled into impulse purchases. It reads product specifications, compares reviews, and matches features to the criteria its user set. The keywords that win a Google auction are irrelevant to an AI that parses structured product data. For brands that have spent decades perfecting their search marketing, this is a fundamental change in what “being found” means.
Product data becomes the new marketing asset. The specifications, descriptions, and structured attributes in your product feed are what the AI reads. Sloppy metadata, missing attributes, or inconsistent formatting mean the agent may never consider your product. This is the argument from Part 1 applied to commerce: the machine reads what you publish, so what you publish had better be clean, complete, and accurate.
And yet brand still matters. An Amperity survey cited by eMarketer found that while 54.5 percent of consumers said a strong AI recommendation could influence their purchase, and more than half said AI had introduced them to brands they found satisfying, only 6.3 percent said they trust AI recommendations more than their own brand preferences. AI is widening the top of the funnel. It is not replacing loyalty at the point of purchase.
What you can do now
Three moves matter right now.
First, audit your product catalog for machine readability. Your product pages were designed for humans who browse. AI agents parse structured data. If your product feeds have gaps, inconsistent attributes, or descriptions written for search keywords rather than factual accuracy, fix them. This is the single highest-leverage change most brands can make, and it applies whether you sell through your own site or through marketplaces.
Second, decide whether you are building your own AI shopping experience or relying on someone else’s. Walmart is developing its own shopping agents and embedding its Sparky chatbot into ChatGPT. Smaller brands may choose to make their catalogs available through Shopify’s Agentic Plan or through the ACP and UCP protocols instead of building their own. There is no single right answer, but there is a wrong one: doing nothing while the channels form around you.
Third, recognize that for the foreseeable future you are persuading two audiences simultaneously. The AI agent decides which products make the shortlist. The human decides which one to buy. Your structured product data needs to satisfy the algorithm. Your brand story still needs to satisfy the person. Companies that optimize for only one of those audiences will lose the other.
The middleman is here
A new intermediary has entered the relationship between you and your customer. This new reality can work for you or against you depending on the choices you make now.
This series has followed AI’s insertion between brands and customers across three layers. In Part 1, it was about being found: making sure AI engines can read and cite your content. In Part 2, it was about setting terms: deciding which bots get access and at what price. Here in Part 3, it is about being chosen: keeping your brand relevant when an AI agent curates the options.
The through-line is the same at every layer. A new intermediary has entered the relationship between you and your customer. It is not going away. The companies that treat it as a channel to be managed rather than a threat to be resisted will have the advantage, and that advantage will compound as AI adoption accelerates. The new middleman is here. The question is whether it is working for you or around you.
Pfanner Advantage works with clients to turn change into advantage at the intersection of mobility, motorsport, media, technology, and marketing. Learn more or start a conversation: contact us today.
More Cold Read Insights from Bill Sparks:
Should You Allow AI Bots to Crawl Your Site
How AI Is Getting Between Brands And Customers
What 1979 Tells Detroit About 2026
What Aren’t Americans Buying EVs?
Part 8: State of the Art Branding Case Studies
Part 7: How the Ad Industry is Reinventing Itself
Part 6: AI and the Personalization of Brand Narrative
Part 5: Building Brand Loyalty Through Community
Part 4: Independent Creators as Brand Storytellers
Part 3: Brands as Publshers: The Rise of Owned Media
The Advantage Journal arrives every week. What matters in sport, mobility, media, and technology — curated and contextualized by Bill Sparks, Bill Long, and Paul Pfanner. No hedging. No filler. Subscribe — It’s Free
