June 24, 2026. For most of the last two years, sending an AI assistant to do your shopping was a novelty that rarely ended in a purchase. That has flipped. New data from Adobe Digital Insights, drawn from more than one trillion visits to United States retail sites, shows that traffic arriving from AI assistants and AI powered browsers now converts better than traffic from any other channel. In March 2026, shoppers referred by AI converted 42 percent better than non-AI visitors, a record high and a sharp reversal from a year earlier, when the same traffic converted 38 percent worse. The AI shopper is no longer just browsing. The problem for most merchants is that their store was never built to be read by a machine.
The shift in the numbers
Adobe's latest retail report puts hard figures behind a trend operators have felt for a year.
- AI traffic to United States retail sites grew about 393 percent year over year in the first quarter of 2026 (January to March), after climbing roughly 693 percent year over year during the November to December 2025 holiday season.
- In March 2026, AI referred traffic converted 42 percent better than non-AI traffic, up from converting 38 percent worse in March 2025.
- These shoppers also engage more once they arrive: 12 percent higher engagement, 48 percent longer on the site, and 13 percent more pages per visit.
- Trust is rising. 39 percent of shoppers say they have used AI to shop online, 85 percent of them say it improved the experience, and 66 percent believe AI tools give accurate results.
- The money is already large. Salesforce estimated AI and agents influenced about 262 billion dollars, roughly 20 percent of global online holiday spend, in the 2025 season, and McKinsey projects agentic commerce could drive 1 trillion dollars in United States retail revenue by 2030.
Why most stores are invisible to AI
If AI shoppers convert so well, why is not every merchant winning? Because an AI assistant can only recommend, compare, or buy a product it can actually read, and Adobe's benchmark shows most retail content is only half legible to a machine. Adobe scored retail pages with an AI visibility checker, where the score is the share of a page a model can read. Across United States retail, homepages average 75 percent and category pages 74 percent, but individual product pages, the pages that decide a sale, score just 66 percent. The reason is familiar: specs, sizing, materials, and compatibility are often locked inside images, scripts, or unlabeled markup that an assistant cannot parse. Support pages actually score higher (returns 82 percent, contact 81 percent, FAQ 80 percent, help center 79 percent), which means the weakest content is exactly where the buying decision happens. The gap between leaders and laggards is wide, from 82.5 percent on the best homepages down to 54.2 percent on the worst. When an agent cannot read your product page, it does not wait. It moves to a competitor it can read.
Where the AI shoppers actually are
The traffic is arriving on two fronts, and you need to be present on both. On the open web, assistants now buy through emerging standards: OpenAI's Agentic Commerce Protocol with Stripe powers checkout inside ChatGPT, Google's Universal Commerce Protocol connects agents to merchant catalogs, and Microsoft Copilot and Perplexity are building their own shopping flows. Shopify merchants can syndicate to these surfaces through Agentic Storefronts from a single admin. We covered the discovery first reality of this channel, after ChatGPT pulled back its Instant Checkout pilot, in our June 12 analysis. On the other front sits Amazon's walled garden. Its Alexa for Shopping assistant, which absorbed the Rufus assistant that helped more than 300 million customers in 2025, now tracks price targets, builds carts on a schedule, and through a Buy for Me feature can complete a purchase on the shopper's behalf, even on other retailers' sites. The lesson is not to pick a side. It is that your catalog has to be present and readable in both the open channels and the closed ones.
What it means for operators
You do not need an enterprise budget to get ready. Most of the work is hygiene, and it compounds. Building on the readiness plan in our June 10 Shopify playbook, here is the short list that matters now.
- Make product pages machine readable. Put specs, sizing, materials, and compatibility in real text rather than baked into images, add Product structured data, and use clear titles and plain language. Run a page through an AI visibility checker and fix what a model cannot see.
- Keep your product feed clean and syndicate it. Accurate titles, prices, availability, and images are what the agent channels read. On Shopify, switch on the agentic surfaces. On other platforms, keep your merchant feeds current.
- Write the boring pages clearly. Shipping, returns, and policy pages are what an agent checks before it recommends you, so make them unambiguous.
- Own the customer, not just the sale. Capture first-party data and keep your own checkout strong, so when an agent sends a buyer you can serve and re-engage them rather than renting the relationship.
- Measure AI traffic on its own. Segment AI referrals in analytics so you can see what converts and where to invest.
This is the work that pays off before the next holiday season, when AI referred traffic already jumped close to 700 percent a year ago and will be larger again. We help merchants get their storefronts, feeds, and product data ready for AI driven discovery and checkout through our ecommerce and Shopify app work, you can hire a Shopify expert to do the migration, and pairing it with AI automation keeps product data clean at scale. The merchants who win the AI shopper in 2026 will be the ones a machine can actually read.
Frequently Asked Questions
It means the content an AI assistant needs, product specs, prices, sizing, shipping and return terms, exists as real text and structured data rather than being locked inside images, scripts, or unlabeled markup. Adobe found average retail product pages are only about 66 percent machine readable, so a third of the deciding content is invisible to agents.
Yes. Adobe reports AI referred traffic converted 42 percent better than non-AI traffic in March 2026, a record and a reversal from converting 38 percent worse a year earlier, and those shoppers stay 48 percent longer and view 13 percent more pages. The channel is small but growing fast and converts above average.
Keep an accurate product feed, switch on Shopify's agentic surfaces so your catalog syndicates to ChatGPT, Google, Copilot, and Perplexity, add Product structured data, and write clear product, shipping, and return pages. Then test a few key pages with an AI visibility checker and fix anything a model cannot read.
No. Amazon's Alexa for Shopping is one large channel, but it is a closed one. The open channels, ChatGPT, Google, Copilot, and Perplexity, reach shoppers buying across the web. The goal is to be present and readable in both rather than betting your store on a single platform.
Your product pages. They convert the sale yet score the lowest for machine readability. Move specs and key details out of images into real text, add Product structured data, and make titles and descriptions explicit. That one change helps both AI assistants and traditional search.
Almost certainly. AI referred retail traffic rose roughly 693 percent year over year last holiday season and has kept climbing through 2026. The stores that are machine readable and syndicated to the agent channels before the season will capture demand the laggards cannot.