July 15, 2026. For two years, e-commerce operators could file AI shopping under "interesting, not urgent." The visitors arriving from ChatGPT, Gemini, Copilot and Perplexity were a rounding error, and when they did land, they converted worse than almost any other channel. Adobe's final accounting of the June Prime Day event, published June 29, says that second part is now officially over. AI referred shoppers converted 40 percent better than non-AI channels. A year ago they converted 23 percent worse. That flip, plus a July calendar full of agent storefront launches from Salesforce, Shopify, Google and Microsoft, is why peak season planning looks different this year.
What the data says
The numbers below come from Adobe's final Prime Day report (published June 29, covering June 23 to 26 across more than 1 trillion visits to US retail sites) and from Salesforce's June 24 Agentforce Commerce release, with the wider race covered the same week by PYMNTS.
- The conversion flip is real. During the Prime Day event, visitors who clicked through to US retail sites from generative AI tools converted 40 percent better than non-AI channels such as paid search, email and social. At the same event in 2025, AI traffic converted 23 percent worse.
- AI referral volume is compounding. Adobe measured AI traffic to US retail sites up 235 percent year over year across January to May, and up 89 percent during the event itself.
- The event set a record. US shoppers spent 26.4 billion dollars online from June 23 to 26, up 9.3 percent year over year and ahead of Adobe's own forecast. Mobile took 54.2 percent of sales, an all-time high, and buy now pay later covered 2.1 billion dollars of it.
- Retailers running their own shopper agents are pulling ahead. Salesforce's holiday 2025 data, restated in its June 24 release, put AI influenced sales at 20 percent of global online revenue, worth 262 billion dollars, and found retailers operating their own shopper agents grew sales 59 percent faster than those without.
- The agent storefront calendar is concrete now. Salesforce made its Shopper, Buyer and Merchant agents generally available on June 24, with its ChatGPT product channel scheduled for GA in July and Google Search, AI Mode and the Gemini app to follow this summer. Shopify merchants already have Agentic Storefronts reporting in admin and the UCP based cart infrastructure we covered last week.
- Back to school is already spiking. During the Prime Day event, kids apparel ran 140 percent above average June daily sales, backpacks 115 percent and lunch boxes 105 percent. The next shopping season did not wait for September.
Why the flip happened
The mechanics are not mysterious. A shopper who asks an assistant "best budget espresso machine under 300 dollars with a steam wand" has already done the comparison work by the time they click through. They arrive on a product page pre-qualified, often with the exact model chosen. Adobe's own framing is that AI tools are shortening the time it takes consumers to confidently get the information they need. The same pattern showed up on the B2B side in Forrester's buying journey data: research starts in the assistant, the click happens near the decision.
That changes what your site needs to do. An AI referred visitor does not need your quiz, your popup or your category carousel. They need the product page to load fast, confirm stock, state the delivery date, show the total landed cost and offer the payment flexibility they expect. The conversion is yours to lose after the click.
The channel calendar you can act on this month
The strategic question has moved from "should we care about AI shoppers" to "which surfaces can we switch on before peak." As of this week: Shopify merchants can review their AI channel presence under Agentic Storefronts in admin and keep product data clean for the catalog feeds that power ChatGPT and Copilot surfaces. Salesforce B2C merchants can deploy Shopper Agent on their own storefront today and connect catalogs to ChatGPT through Business Manager as the channel goes GA this month, staying merchant of record. Google's agentic checkout and Universal Cart continue rolling across Search, AI Mode and Gemini this summer. Microsoft's Copilot Checkout is live for participating merchants. None of this requires abandoning your own storefront. The point, as Salesforce's commerce lead put it in the June 24 release, is that referral will be AI influenced while transactions still land predominantly on owned properties.
Back to school is the rehearsal for holiday
Back to school is now the first full season where AI referral is a measurable channel, and it is running. If your store sells anything adjacent to the category spikes Adobe recorded, the window to show up in assistant answers is the next four to six weeks. Holiday is the same exercise with higher stakes: most platforms enter code freeze and promo lock windows in October, which means the machine readability work, the feed hygiene and the agent deployments need to happen in August and September. We walked through the readiness checklist in our June 24 merchant readiness playbook; the Prime Day data is the evidence that the checklist pays.
What it means for operators
A six step sprint between now and Labor Day, in priority order.
1. Baseline your AI referrals this week. Segment traffic from chatgpt.com, gemini.google.com, copilot.microsoft.com and perplexity.ai in analytics. You cannot manage a channel you do not measure, and you want a before picture.
2. Make every product page machine readable. Complete Product schema, real availability, visible pricing including duties for cross border buyers, and plain language spec tables. This is the highest leverage work per hour spent, and it is exactly the kind of project a Shopify expert can knock out in a sprint.
3. Turn on the channels your platform already ships. Shopify Agentic Storefronts, Salesforce Agentic Commerce Channels, Copilot and Google feeds. Inventory what is one toggle away before buying anything new.
4. Put an assistant on your own storefront. Salesforce's 59 percent figure is vendor reported, but the direction matches what Adobe measured: stores that answer questions convert decided visitors better. On Shopify, start with intelligent product page FAQs before a full agent.
5. Fix the post-click path for decided buyers. Speed, stock clarity, delivery dates, BNPL at checkout. AI shoppers traded up to pricier goods during Prime Day, and BNPL grew 9.5 percent; the two facts belong together.
6. Plan promos AI can see. Competitive discounts drove the record, with electronics at 24 percent off. Assistants compare prices relentlessly, so plan peak discounts you can sustain and state them in structured data. If you want the full e-commerce readiness work handled end to end, that is what we do.
The honest caveats
Adobe itself notes AI referral volume remains modest next to search and email, so this is a growth channel, not your revenue base. Adobe measures clicks to retail sites, not purchases completed inside assistants, and Salesforce's figures describe activity across its own merchant network. Conversion lift also is not incrementality: some AI referred buyers would have found you anyway. The reason to move now is not that AI referrals dominate today. It is that the channel doubled in a year, flipped from worst to best converting, and the platforms just shipped the rails, which is the same setup organic search had in 2005. On the paid side, the budget is already following the same path, as we covered in yesterday's ChatGPT Ads piece.
Frequently Asked Questions
Adobe's final report, published June 29, 2026, found visitors arriving at US retail sites from generative AI tools converted 40 percent better than non-AI channels during the June 23 to 26 event, a reversal from 2025 when AI traffic converted 23 percent worse. AI traffic to US retail sites was up 235 percent year over year from January to May and up 89 percent during the event.
Not yet as a share of revenue, and Adobe says so plainly: volume remains modest next to search, email and social. It matters because of trajectory and quality. The channel roughly doubled in a year, the visitors it sends are the best converting Adobe measures, and the work to capture it (structured data, clean feeds, fast product pages) also improves regular SEO and conversion.
Shoppers increasingly do their comparison work inside the assistant, so the click happens near the decision rather than at the start of research. Adobe credits AI tools with shortening the time it takes consumers to confidently get the information they need. Better product data on retail sites over the past year also gave assistants more accurate answers, so fewer clicks land on the wrong page.
Three things in order: segment AI referral traffic in analytics so you have a baseline, complete Product schema and availability data on your best selling product pages, and review the Agentic Storefronts section in Shopify admin to see how your store already appears across AI channels. After that, tighten the post-click path: page speed, delivery dates and payment options for buyers who arrive decided.
Treat them as directional. The 59 percent and the 8x social conversion figures come from Salesforce's analysis of activity across its own merchant network, and vendors publishing stats that flatter their products deserve scrutiny. The reason to take the direction seriously is that Adobe, an independent measurement source, found the same pattern: AI touched journeys convert unusually well.
Work back from platform freeze windows. Most retail teams lock site changes and promo calendars in October, so machine readability, feed hygiene and any storefront agent deployment realistically need to ship in August and September. Back to school is the live rehearsal: it is the first full season with AI referral as a measurable channel, and it is running now.