On August 7, Reuters published a report that PYMNTS also covered immediately. The anchor data point: 49.6 million U.S. adults, roughly 19 percent of the population, now begin their retail product search with AI. Among those, 39 million use AI as their primary starting point. The other 10.6 million use it alongside traditional channels. Juniper Research estimates those shoppers will drive $8 billion in retail spending this year after AI agents direct them to retail sites. Adobe Analytics found that AI-referred visitors generated 41 percent more revenue per visit than shoppers arriving through any other channel.
The discovery question is already answered. AI is the new top of the retail funnel for a significant and growing share of the consumer market. The story still being written is what happens after the discovery. A shopper finds a product through Gemini or ChatGPT. They click through. They land somewhere. The question every retailer with a presence in the AI discovery layer is now solving in real time is straightforward. Does that shopper complete the transaction on the retailer’s own platform? Or does the friction of the redirect send them back to the AI interface to buy from whoever is transactable inside the conversation?
Showing up in the AI search result is the new SEO. It is necessary. It is not sufficient. The retailers who win the next five years of AI-driven commerce solve the discovery layer, the transaction layer, and the data layer simultaneously. Three retailers explained this week exactly how they are doing it.
The Tax That Never Goes Away
Josh Friedman, head of digital and eCommerce at Ulta Beauty, put the structural reality in terms every retailer should recognize: “Whether it’s Google search, affiliate marketing or Facebook, there’s always a tax for engaging customers on other people’s platforms. I don’t think this is any different.”
That framing removes the novelty from the AI commerce conversation and puts it in a context retailers have managed for twenty years. The platform tax is not new. Google search created it. Affiliate marketing formalized it. Social commerce extended it. AI discovery is the latest version of the same structural tension: the customer finds you somewhere you do not own, and you pay a cost, in data, in margin, or in relationship control, to be present there.
What is different this time is the scale of what is at stake. As I described in the AI-referred traffic conversion analysis, AI-referred shoppers convert at higher rates and generate more revenue per visit than visitors from any other channel. The platform tax on AI discovery is therefore more expensive than any previous version of it. The customers you are paying to reach are your most valuable ones.
What Ulta Beauty Got Right
Ulta is incorporating its shopping cart and loyalty program directly into Google Gemini. Customers find Ulta products in Gemini results. The transaction preference, however, is clear: Friedman wants customers to complete the purchase on Ulta’s own website. The reason is not checkout friction or UX control. It is the loyalty program. Ulta has built one of the most valuable loyalty databases in U.S. retail. Every transaction that completes on Ulta’s own platform feeds that database. Anything that completes inside Gemini does not. The integration with Gemini is a discovery investment. The redirect to Ulta’s own checkout is a relationship investment. Those are two different strategies working together.
What The Knot Got Right
The Knot is optimizing its website so its vendors appear in ChatGPT results. Raina Moskowitz, CEO of The Knot, explained the transaction preference directly: “We have three decades of data to help” customers coordinate weddings. That data lives on The Knot’s own platform. A customer who books a venue through ChatGPT loses access to thirty years of wedding planning intelligence. The Knot’s argument for its own checkout is not price or convenience. It is value that only exists on its own platform. That is a defensible position because it gives the customer a reason to transact on The Knot’s site that no AI interface can replicate.
What Etsy Got Right
Rafe Colburn, chief product and technology officer at Etsy, offered the most precise observation in the Reuters piece. He found that users who discover products through ChatGPT tend to complete purchases on Etsy’s website anyway. His conclusion: “It’s the start of a deeper relationship, and not just one that’s always intermediated by ChatGPT.” Etsy’s product is fundamentally about the relationship between a buyer and a maker. That relationship requires Etsy’s own platform to exist. The AI discovery is the introduction. The transaction is the handshake. The relationship is what happens after. For Etsy, the AI channel deposits high-intent customers at the top of a relationship funnel the AI cannot replicate.
The Three-Layer Problem Every Retailer Has to Solve
Vince Koh, global head of digital commerce at AWS, summarized the strategic imperative: “When a shopper completes a purchase on a brand’s own site, the retailer maintains a direct relationship with that customer.” AWS now actively advises its retail clients to appear in AI platforms for discovery while ensuring shoppers transact on their own websites. That is not a defensive posture. It is a recognition that AI commerce operates on three layers simultaneously. Retailers who optimize for only one will lose the other two.
The Discovery Layer: Are You Visible to the Agent?
This is the GEO problem I described in the Generative Engine Optimization analysis. Three inputs determine whether a product appears when a shopper asks an AI agent for a recommendation. The first is product data built for natural language queries. The second is editorial authority in sources that informed AI model training. The third is catalog presence in AI commerce networks. Retailers who have not solved this layer are invisible to the 49.6 million Americans who start their search with AI.
The Transaction Layer: Does the Click Lead Somewhere Worth Going?
A shopper who finds a product through ChatGPT or Gemini and clicks through to a retailer’s site makes a decision at the moment of redirect. If the landing page is slow, unoptimized for mobile, or requires account creation before purchase, the shopper bounces. They return to the AI interface. They buy from whoever is transactable inside the conversation. As I described in the ChatGPT cart and Shoppable Universal Checkout analysis, the friction of the redirect is the primary conversion killer in AI-driven commerce. The retailers who win this layer have built landing experiences designed for AI-referred intent, not for organic search intent. Those are different visitors with different context and different expectations.
The Data Layer: Is the Transaction Building Something?
This is the layer most retailers have not yet designed for. When a transaction completes on the retailer’s own platform, it generates data. That includes purchase history, browsing behavior, loyalty points, cross-sell signals, and the attribution record showing this customer arrived from an AI referral. That data feeds future personalization, future loyalty offers, and future AI recommendations that surface the retailer’s products again. When the same transaction completes inside a third-party AI interface, the retailer gets the order. The platform keeps the data. Ulta’s loyalty program, The Knot’s coordination database, and Etsy’s relationship model are all versions of the same answer. Give the customer a reason to complete the transaction on your own platform. Make that transaction worth more there than anywhere else.
What This Means for LatAm Retailers
The three-layer problem is not a U.S.-only challenge. WhatsApp commerce in Latin American markets already operates on the same dynamic. A shopper in São Paulo or Bogotá who asks Meta AI for a product recommendation starts at the top of that same funnel. The discovery layer belongs to Meta. The transaction layer belongs to whoever makes the redirect worth completing. LatAm retailers with strong loyalty programs, proprietary customer data, or service models unique to their own platforms are well positioned to address the data layer problem. The ones who have not built those assets have less to defend when the customer discovers them through AI and decides the AI interface is simply easier.
Furthermore, the 41 percent revenue-per-visit premium that Adobe documented for AI-referred shoppers applies equally to LatAm markets as AI-assisted shopping scales through the channels those consumers already use. The retailers who invest in the transaction and data layers now will have a conversion infrastructure that compounds. The ones who wait will be optimizing their redirect experience under competitive pressure instead of under strategic choice.
The Honest Question About Your Own Checkout
Ulta, The Knot, and Etsy each solved the three-layer problem from a different starting point. Ulta solved it through loyalty integration. The Knot built its case on proprietary data value. Etsy’s answer was relationship architecture. The common thread is that each retailer gave the AI-referred customer a reason to complete the transaction on its own platform. That reason was specific to what each retailer owns. Nobody else can replicate it.
Most retailers have not yet asked themselves what that reason is for their own customers.
What the AI Interface Cannot Replace
When a shopper finds your product through ChatGPT or Gemini and clicks through to your site, what does your platform offer that the AI interface does not? If the answer is a faster checkout, a loyalty point, and a confirmation email, that is not a defensible position. The AI interface will eventually offer all three. The retailers who survive the platform tax are the ones whose own platform is worth the redirect. What is on the other side of that click has to be genuinely better than what the AI could have provided inside the conversation.
49.6 million Americans already start their retail searches with AI. The discovery layer is not coming. It is here. The retailers who show up in it and give customers a reason to complete the transaction on their own platform will define the next decade of retail loyalty. The ones who show up and lose the transaction to the interface will have paid the platform tax twice: once for the traffic, once for the customer they did not keep.
If you are building your AI commerce strategy or evaluating your transaction and data architecture for AI-referred traffic, connect with me here or reach me on LinkedIn. I am happy to walk through the framework we use across the U.S. and Latin America.
Adriana Rivas is a retail technology executive and AI strategist, and the founder of a U.S.-based hardware company specializing in self-service kiosks, POS systems, electronic shelf labels, and digital signage deployed across the United States and Latin America. She is the award-winning author of How to Implement Self-Service Without Failing (Amazon #1 Hot New Release, Silver Nonfiction Book Award 2025) and recipient of the Gold Stevie® Award, Thought Leader of the Year 2026. She is also recognized by Thinkers360 as the #7 Global Thought Leader in Retail and a Certified Master Expert in Retail.