AI agents surpass human traffic: how retailers can prepare

TL;DR

AI agents now account for more web traffic than humans, per Adobe, and the shoppers they refer convert far better than other visitors. Retailers that publish clear, text-based product information and test their sites with real agent tasks have a better chance of being recommended this holiday season.

Key takeaways

AI agents and crawlers surpassed human web traffic in June 2026, according to Adobe.
AI-referred retail traffic converted 90% better than non-AI traffic in September 2026 and generated 83% more revenue per visit, according to Adobe figures reported by Small Business Trends.
Adobe's AI Content Visibility Checker gave U.S. retail homepages an average score of 75% in April 2026, leaving about a quarter of their content unreadable to AI.
Prices, availability and return rules should live in readable text, not in images, so agents can verify them.
Businesses can audit their site by asking an AI tool to complete a real task from discovery through checkout.
Clear approval points before payment and reliable inventory data protect against agent mistakes.

AI agents and crawlers surpassed human traffic on the web for the first time in June 2026, according to Adobe. By September 2026, AI-referred shoppers had started converting far better than traditional visitors, and the company is telling retailers of every size to prepare their websites for a new kind of customer: one that generally needs to verify facts and complete a specific task rather than be persuaded by branding or storytelling.

For a small retailer, the change is direct. A shopper may soon hand the whole research task to an agent, comparing prices, checking delivery dates and completing checkout with one approval. A website may hide that information in images or broken controls. The agent then simply moves on to a competitor. The owner may never know a motivated buyer visited at all.

AI traffic is growing and spending more

The scale of the shift comes from Adobe’s analysis of more than 1 trillion visits, 100 million products and 18 retail categories. AI traffic to U.S. retail websites grew 71% year over year in September 2026, according to Adobe figures reported by Small Business Trends. AI-driven traffic to retail sites had already grown 127% year over year in August 2026. The August figure comes from Adobe’s Holiday Shopping Trends 2026 report, while the September figure is from a separate Adobe analysis of over 1 trillion visits.

What those visitors do once they arrive matters more than the volume. In September 2026, AI-referred traffic converted 90% better than non-AI traffic. It generated 83% more revenue per visit. It also added products to carts 53% more often. Just one year earlier, in August 2025, AI visits converted 9% worse than non-AI visits. The reversal reflects how people now use AI tools to narrow their choices before clicking through: by the time they land on a retailer’s site, much of the research is done and the intent to buy is real.

What an AI agent actually checks

Adobe tested an agent with a real request: a three-day Ha Long Bay cruise with kayaking for less than $700. The agent selected four options, then used a browser assistant to check availability, room types, activities and prices on booking pages. It verified facts and completed a specific task, unlike a human visitor who might be moved by branding or storytelling.

That difference exposes weak points a business may not notice. A price embedded only in an image is hard for software to read. An unlabeled date picker can prevent an agent from selecting travel dates. A page that shifts as it loads disrupts automated browsing, and a bot protection system may block a legitimate agent along with malicious traffic. A human who cannot complete a booking might call or leave a trace in analytics; an agent may simply recommend a competitor.

Adobe describes two stages in the selection process. An agent first decides which brands to consider through search and other information sources. It then evaluates the websites it visits. A business therefore needs both a discoverable online presence and a site agents can use once they arrive. Slow or unstable pages may cause agents to abandon a site without leaving a visible trace.

Whether AI can read your homepage decides if it recommends you

Adobe’s AI Content Visibility Checker, which scores how much of a page large language models can read, gave U.S. retail homepages an average of 75% in April 2026: about a quarter of their content was invisible to AI. The best-performing retail sites scored 82.5%, the weakest 54.2%. Promotional and navigation-heavy layouts provide less usable context.

The fix starts with the information a business already publishes. Product pages should state prices, sizes, materials, availability, delivery windows, return rules and restrictions in readable text, never only in images or animated elements. A local clothing store can add sizing guidance and fabric details; a home goods business can explain installation requirements or compare products by room size. FAQs on shipping deadlines, warranties and payment options help both customers and AI systems.

More text is not automatically better, though. Descriptions copied from suppliers, outdated stock information or vague claims can confuse shoppers and damage trust if an AI system repeats them.

Testing your site before the holiday season

Adobe’s practical advice is to audit the site the way an agent would. Ask an AI tool to perform a real customer task, find a product under a set budget, book an appointment or compare shipping options. Track whether it can find the relevant page, understand the offer, use the controls and reach checkout. Repeat the test on phone and desktop, and check it with keyboard navigation and screen-reader tools.

The basics remain familiar: descriptive form labels, accessible menus and fast, stable pages. Adobe also points to opportunities after the first sale. Agents can follow up on unfinished bookings. They can remind customers to reorder consumables or prompt clients to schedule seasonal maintenance.

Owners should still move carefully. Agents can misunderstand product details or act on outdated information. Businesses therefore need clear approval points before payment. They also need safeguards around customer data and reliable inventory and pricing systems. Adobe promotes its own tools for this work, such as Experience Manager Sites Optimizer and Adobe Brand Visibility, and points to WebMCP, a proposed web standard that gives agents a direct way to take actions on a site. But smaller companies may get most of the benefit from structured content, accessibility fixes, faster hosting and manual agent testing. The work does not end with one review: prices, page templates and AI capabilities will keep changing, and monitoring agent-driven visits and failed task journeys may show early where customers are being lost.

Sources: