Shoppers now ask AI what to buy. Traffic from AI answers to retail sites grew 1,324 percent in nineteen months.

The traffic is real and compounding

Adobe Analytics, which began tracking AI referrals in October 2024 across more than a trillion visits to US retail sites, reported in March 2025 that traffic from generative AI sources had jumped 1,200 percent since July 2024. By June 2026 the picture had hardened: AI-referred traffic was up 138 percent year on year in May 2026 and up 1,324 percent since tracking began.

The behaviour of those visitors changed faster than the volume. In Adobe's early data, AI-referred shoppers converted at barely half the rate of other traffic: they were researching. By mid 2026 the relationship had reversed, with AI-referred traffic converting 54 percent better than non-AI sources, spending 53 percent more time on site and browsing 23 percent more pages. The assistant now does the comparison work before the click, so the visitor who arrives has largely decided.

The holiday season put a commercial scale on it. Salesforce's 2025 holiday data, published January 2026, put global online sales at $1.29 trillion and attributed $262 billion, around 20 percent, to AI-influenced journeys, with traffic from AI search channels roughly doubling year on year and those shoppers converting nine times more often than visitors from social referrals. Triple Whale's 2025 benchmark across 33,000+ brands measured AI-driven orders up 1,481 percent for the year, with ChatGPT taking 97 percent of that share. The honest caveat: growth rates this large sit on a small base, and none of these sources publishes AI referrals as a share of total traffic, which for most brands is still low single digits. The direction is not in doubt; the current size is easy to overstate.

What actually changes for a D2C brand

The assistants do not browse your homepage or admire your brand film. They read text, follow structure and weigh corroboration, and they compress your category into a short answer. Three things follow.

• Product pages become source material. An engine assembling an answer needs facts it can lift: what the product is, who it is for, sizes, ingredients or materials, price, shipping and returns, stated in plain text with proper structured data rather than locked in images or a founder story. A page written only to persuade a human in the moment gives the machine nothing to quote. This is Landing and sales pages work with a second audience in mind, and the two audiences mostly want the same clarity.
• Reviews carry more weight than you planned for. The engines lean on third-party corroboration: review platforms, forums, comparison articles, editorial lists. A brand whose proof lives only on its own site is invisible to a system built to distrust self-description. Review volume, recency and the specific products reviews mention become discovery assets, not just conversion assets.
• Comparison content answers the actual question. Shoppers ask assistants comparative questions: best X for Y, this brand versus that, is it worth it. The sources engines cite for those answers are pages that answer them honestly. If nobody has written the comparison your buyer asks for, the engine will find someone else's version with someone else's conclusion.

There is also a harder structural point. A search results page shows ten choices and lets a weak brand win a click with a good title. An assistant's answer names two or three, and everyone else is absent. That compression rewards the brands the engines can verify, name consistently and corroborate off-site, and it punishes brands whose facts differ between their site, their retail listings and the directories. Consistency of the basics, name, range, pricing claims, stockists, across everywhere the engines read is dull work with a direct payoff here.

The assistant does not leave at checkout

The same Salesforce dataset shows the assistants working after the sale as well as before it: agentic customer conversations jumped 66 percent in December 2025 over November, and AI agent use in customer service ran 126 percent above the prior two months during the holiday rush. For a small team this is the least glamorous and most immediately practical end of the trend, because service load, order status, returns and sizing questions, is exactly what drowns an owner-led brand at peak. The discovery shift gets the headlines; the service shift is already operational.

Measurement before opinion

Most analytics setups file AI referrals under direct or generic referral traffic, so the first step is unglamorous: segment the known AI referrers in your analytics, watch the conversion rate of that segment against the site average, and record which pages they land on. Given Adobe's finding that these visitors convert well, a brand seeing meaningful AI-referred sessions already has evidence worth acting on; a brand seeing none has a baseline. Either result is more useful than a hunch. And because an AI-referred visitor may arrive once and never see your ads, capturing them onto the list on that first visit matters more, which is where CRM, email and retention meets this subject.

What to do

Run your own category questions through the major assistants and record whether you are named, mentioned or absent, and which sources are cited. Fix the product pages the engines would need to quote. Build review depth where the engines actually read. Segment AI referrals in your analytics this month so the trend is yours rather than Adobe's. This is the work our AI visibility, SEO and local search service exists for, and the measurement side sits with Analytics and reporting; the free Business Scan includes the named-or-absent check as standard.