Sources: Adobe Digital Insights, July 2026; ReFiBuy x Digital Commerce 360, Q2 2026 AI1000; Salesforce 2026 holiday predictions.
AI referrals are converting 60% higher
AI-referred shoppers are now retail's highest-converting traffic. Not the biggest source, and not the loudest, but the traffic that converts into a purchase at the highest rate.
Adobe's July 2026 data makes the case. Traffic arriving at U.S. retail sites from AI sources converts 60% higher than non-AI traffic. July marks the eleventh consecutive month AI-referred traffic has held that lead. This is not a spike or a novelty effect. It is nearly a year of sustained behavior.
Adobe's engagement data explains the conversion numbers. Once AI-referred shoppers land on a U.S. retail site, they add items to their cart at a 28% higher rate, spend 59% more time on the site, and bounce 33% less often than shoppers arriving from other channels. They behave like people who have already made a decision, because they have. The volume is building behind them: Adobe measures AI traffic to U.S. retail sites up 62% year over year, and up 1,219% since it began tracking in October 2024.
Retail has a well-worn playbook for valuing a channel: engagement, conversion rate, growth. A channel that demonstrates measurable ROI by turning general queries into qualified, converted buyers is the channel that will own investment. Historically, paid search and SEO earned their budget that way. Email retention programs did too. Judged by that same playbook, AI-referred traffic has spent eleven consecutive months out-converting non-AI traffic, demonstrating that it is exactly the channel this playbook was built to find.
The data indicates something specific: shopper behavior has already shifted, and retail readiness has not caught up. Shoppers are researching and deciding inside AI experiences, then arriving at retail sites ready to buy. Adobe’s average retail homepage visibility score was 61%, which Adobe says means nearly 40% of homepage content was not fully readable by machines. That distance between how shoppers now buy and how prepared retail is to be found is the opportunity. This is new territory, and most commerce teams are still working out how it operates. The rest of this piece is about that gap, what drives it, and who is closing it.
The reality: most retail sites are not ready for AI shopping
Adobe's data describes what is arriving at retail's front door: traffic that converts, engages, and grows. What it cannot describe is what happens behind the scenes, whether a retailer's catalog is actually prepared for the systems doing the shopping. That is the question the AI1000 exists to answer.
The AI1000, the quarterly benchmark ReFiBuy co-developed with Digital Commerce 360, scores the Top 1000 retailers by online sales on agentic commerce readiness: whether each catalog is accessible, visible, gaining momentum, and positioned to compete as AI shopping evolves. The Q2 edition shows a field that is not keeping pace with the channel. The index average fell to 39.7 from 42.0 in Q1, meaning the readiness bar is rising faster than most retailers are clearing it. Only 225 of the 1,000 have a detectable Universal Commerce Protocol endpoint, the emerging standard that lets AI agents connect directly to a catalog. And 972 of the 1,000 changed rank in a single quarter, evidence of how unsettled readiness still is across the field.
Unreadiness carries a specific cost in this channel. Pricing, availability, product specifications, the details a shopper's agent needs to include a product in its answer, only count if agents can read them. Product data that agents cannot parse is not penalized or ranked lower. It never enters the conversation.
None of this shows up in brand recognition, marketing budget, or any analytics dashboard a commerce team checks daily. Yet it is deciding who participates in a channel that converts 60% higher than non-AI traffic. The next section explains the mechanism.
The mechanism: how AI shopping actually decides
The consumer journey behind all of this has three steps: research, find, buy. What changed is where the first two happen.
A shopper opens an AI assistant and describes what they need. A trail running shoe for wide feet under $150 that is currently in stock. The agent reads product data across the web, compares options, and returns a shortlist. The shopper picks one and clicks through to buy. By the time a retail site records the visit, the research is done and the decision is made. That is the conversion pattern reflected in Adobe's data: the site is not winning the sale, it is simply converting it.
These are not search engines, and they do not behave like them. A search engine returns links and leaves the hard work to the shopper. An answer engine returns a shortlist, refined to the shopper's criteria and built from the product data it can read. On a results page, position ten still gets seen. In an answer, there is no page two. A product is in the shortlist or it is absent.
This is why AI readiness decides who is eligible. Agents carry no loyalty to brand campaigns or traditional search rankings. They work with what they can read and confidently interpret. A mid-size retailer with clean, structured product data is fully legible to an agent, whatever its ad budget.
The AI shopping leaderboard is still wide open
The Q2 AI1000 shows readiness being scored right now. The benchmark ranks the Top 1000 retailers on agentic commerce readiness, and Q2 made two things clear: the field is wide open, with 972 of 1,000 retailers changing position in a single quarter, and the retailers leading it are the ones that made their product data readable to agents ahead of everyone else. Adobe's category data points the same way: apparel, the category scoring highest on machine readability, is the category most represented at the top of the rankings. The retailers doing the readiness work now are the ones positioned to capture the channel as it grows, and to be ready for the season when it matters most.
The window for that work is defined. Salesforce forecasts that 20% of 2026 holiday ecommerce traffic will come from AI chat agents, and the catalogs those agents will recommend in December are being read now. This is the case forAgentic Commerce Optimization (ACO) as a standing function rather than a one-time project: the readiness bar moves every quarter, and the retailers treating catalog readiness as ongoing work are the ones the channel can see. Retail traffic that converts 60% higher than non-AI traffic is already arriving. Eligibility is being decided now.
