ChatGPT shopping referrals more than doubled year-over-year across the US, UK, Germany, and France in 2026, with some retailers now seeing AI account for up to 25% of referral traffic, according to Bain & Company’s May 2026 report on agentic commerce.
ChatGPT shopping referrals more than doubled in a year. For some retailers, AI now drives 25% of referral traffic. This is not a trend to monitor from a distance. It is a land-grab moment, and the window to move first is open right now.
Shopping referrals from ChatGPT more than doubled year-over-year across the US, UK, Germany, and France, according to Bain & Company’s Agentic AI in Retail report published May 13, 2026. For some retailers, AI now accounts for up to 25% of referral traffic. That’s not a rounding error anymore.
Between 30% and 45% of US consumers already use AI tools for some part of their shopping process, a finding from the same Bain report that should recalibrate how any growing brand thinks about discovery budgets and channel mix. This is not an edge case demographic. This is a third to nearly half of the online buying population already integrating AI into product research, comparison, and deal-finding before a purchase decision is made.
64% of US consumers have either used or are open to using AI to complete a purchase. That is nearly two-thirds of the online buying population signaling willingness to let AI handle more of the transaction. When a majority of buyers are already open to AI-mediated purchasing, the brands that are not optimizing for AI discovery are not playing a forward-looking bet. They are already behind.
The implication for operators is direct. If your catalog is not structured in a way that AI shopping engines can parse, evaluate, and recommend confidently, you are invisible to a meaningful and growing share of potential buyers before the consideration stage even begins. The awareness gap is not a paid media problem. It is a product data problem.
Bain projects fully agentic commerce will reach $300 billion to $500 billion in US revenues by 2030, representing up to 25% of total US ecommerce by that point. That projection covers the full spectrum from AI-assisted product discovery to AI-recommended purchases to, eventually, AI-initiated repeat orders. The first two categories are already measurable in retailer analytics today. The third is emerging.
Morgan Stanley’s parallel estimate puts US agentic commerce at $385 billion by 2030, landing right in the middle of Bain’s range. Two independent research firms converging on the same order of magnitude from separate methodologies adds significant weight to the projection. When organizations that size agree on scale, operators who are waiting for more certainty before acting are benchmarking against a signal that is already clear.
The $300B-$500B figure is not a ceiling. It is a baseline projection based on current adoption curves and infrastructure maturity. If AI shopping infrastructure continues to compound at current pace, and if ChatGPT’s 900 million weekly active users continue converting general usage into shopping queries, the actual figure could exceed those projections. The more conservative read is that agentic commerce will be one of the defining acquisition channels of the second half of this decade.
Three forces are compounding simultaneously, and understanding each one helps operators prioritize which response actions will have the most leverage.
Consumer behavior is shifting at platform scale. ChatGPT now has 900 million weekly active users as of May 2026. When a platform that large adds shopping capabilities, adoption does not require convincing. The audience is already there, using the tool daily for tasks ranging from writing and research to trip planning and customer service. Shopping queries are a natural extension of that behavior, not a separate habit that needs to be built. Users do not download a new app or create a new account. They ask a purchase question in a tool they already trust, and the AI returns product recommendations. The habit formation cost is zero.
Retailers are seeing it in their data. Bain’s companion report, Rewiring Demand Generation in the Age of AI Agents, makes the case that demand generation itself is being restructured. When AI referrals account for 25% of a retailer’s traffic, that retailer starts optimizing for AI visibility the same way they once optimized for Google page-one rankings. The behavioral shift among operators is not driven by theory. It is driven by attribution data showing a channel growing faster than anything else in the mix.
The infrastructure is maturing across every major AI platform. Google AI Mode, Perplexity, and ChatGPT are all building structured product experiences. Google already displays AI answers on 48% of searches as of March 2026, according to SEO.com. Product recommendations are a natural extension of that capability, and each platform is investing specifically in making those recommendations more accurate, more personalized, and more directly purchasable. The infrastructure is not coming. It is here and improving quarterly.
The brands that appear in AI shopping recommendations are the ones with rich, structured product catalogs, and the ones that get filtered out are the ones whose data does not give the engine enough to work with. That is the operational reality this moment creates.
A search engine returns ten blue links for “lightweight running shoes under $150.” An AI shopping agent returns three specific products with prices, images, ratings, and an explanation of why each one matches the query. The three that appear are not the three with the best ad spend or the highest domain authority. They are the three whose product data gave the engine enough confidence to make a specific recommendation.
The average product listing has 5 to 8 attributes. AI shopping agents need 30 or more structured attributes to make confident recommendations. Size, material, use case, compatibility, care instructions, sustainability credentials, weight, dimensions, return policy, fit notes: every missing attribute is a reason an AI might surface a competitor instead. The brands doing $2M-plus who are already investing in catalog enrichment will widen that gap significantly over the next 12 to 18 months.
This is the product data gap Bain’s findings make urgent. It is not enough to have a product page that looks good to a human visitor. The data needs to be machine-readable, attribute-rich, and distributed to every AI channel that consumers are starting to use for shopping discovery. Whether you are doing $500K or $5M, the brands that close this gap now are building a compounding visibility advantage that is genuinely hard to reverse once it sets.
Four actions move the needle on AI visibility and catalog readiness, and none of them require waiting for the market to develop further before they are worth doing.
Track your AI visibility. Most retailers have no idea how they rank in ChatGPT, Google AI Mode, or Perplexity for their key product queries. That is the equivalent of not checking your Google search rankings in 2010. The data exists and the tools to surface it exist. Start measuring where you appear and where you do not before optimizing anything else, because you cannot prioritize what you have not measured.
Audit your product data for AI readability. Pull a sample of your top 20 SKUs and count the structured attributes on each one. If the average is under 15, you have a gap that is costing you recommendations right now. AI shopping agents need attributes like use case, compatibility, materials, sustainability claims, and specific fit or sizing notes to match confidently against purchase-intent queries. Thin data means filtered out results.
Publish AI-optimized product feeds. Structured feeds designed for AI consumption are different from traditional shopping feeds. They need richer attributes, consistent formatting across every SKU, and distribution to the specific channels where AI agents pull product data. A feed that works for Google Shopping in 2022 is not optimized for how ChatGPT or Perplexity surfaces products in 2026.
Monitor competitive positioning. Bain’s data shows this is a land-grab moment. The retailers who optimize for AI discovery now will build the same kind of compounding advantage that early SEO adopters built 15 years ago. Monitoring which queries your competitors appear in, and which ones you do not, gives you the targeting intelligence to close gaps before they compound against you.
Paz.ai does all four. It monitors your rankings across ChatGPT, Google AI Mode, and Perplexity, showing exactly which queries find you and which do not. It enriches your catalog and publishes optimized feeds to every AI shopping channel from one dashboard. You can start with a free AI Readiness Report at paz.ai to see where you stand in 30 seconds.
Paz.ai helps brands monitor, optimize, and publish their product catalogs across every AI shopping channel. Run a free AI Readiness Report on your product page in 30 seconds at paz.ai.
ChatGPT shopping referrals more than doubled year-over-year across the US, UK, Germany, and France, according to Bain & Company’s Agentic AI in Retail report published May 13, 2026. For some retailers, AI now accounts for up to 25% of total referral traffic. That growth rate is not a niche signal. It is a mainstream channel shift happening faster than most operators are tracking in their analytics. If you have not separated AI referral traffic from other organic sources in your attribution reporting, you are likely underestimating how much of your current traffic is already AI-driven.
Between 30% and 45% of US consumers already use AI tools during some part of the shopping journey, from product research to comparison to finding deals, according to Bain’s May 2026 research. An additional 64% have either used or are open to using AI to complete a purchase. That means nearly two-thirds of online buyers are not resistant to AI-mediated shopping. They are willing participants waiting for brands to show up in the AI channels they are already using. The adoption is not future-state. It is present and measurable.
Bain projects agentic commerce will reach $300 billion to $500 billion in US revenues by 2030, representing up to 25% of total US ecommerce. Morgan Stanley independently estimates $385 billion for the same period, landing squarely within Bain’s range. Two major research firms converging on the same order of magnitude from separate methodologies is the kind of corroboration that moves a projection from speculative to operationally significant. Brands planning their 2026 and 2027 channel investment mix should treat agentic commerce as a primary growth channel, not a secondary one.
AI shopping engines rely on structured, attribute-rich product data to make confident recommendations, and brands without sufficient catalog depth get filtered out before the consumer ever sees a result. The average product listing has 5 to 8 attributes. AI recommendation engines typically need 30 or more to match a product confidently against a purchase-intent query. Missing attributes, like use case, compatibility, material, care instructions, or sustainability credentials, are each a reason the engine defaults to a competitor with more complete data. The brands appearing in AI recommendations are not winning because of brand equity alone. They are winning because their product data gives the engine enough to work with.
Start by measuring your current AI visibility before optimizing anything else. Most retailers have no data on how they appear in ChatGPT, Google AI Mode, or Perplexity for their most important product queries, which means they are making catalog and feed investment decisions without knowing where their actual gaps are. A free AI Readiness Report from Paz.ai evaluates a product page in about 30 seconds and shows you where you stand. Once you have a baseline, the prioritization of enrichment, feed publishing, and competitive monitoring becomes straightforward rather than speculative.
This article draws on Bain & Company’s Agentic AI in Retail (May 13, 2026), Bain’s companion report Rewiring Demand Generation in the Age of AI Agents (2026), Retail Times coverage from May 18, 2026, SEO.com’s Google AI search data from March 2026, and Morgan Stanley’s US agentic commerce forecast from 2025. All statistics cited are sourced directly from those reports. Where projections are referenced, both the source organization and the publication date are noted so readers can evaluate the data in context.
