Between July 2024 and February 2025, Adobe recorded a 1,200% increase in traffic to U.S. retail websites from generative AI sources. That figure alone would be notable in any industry.
For anyone tracking the broader shift toward automation and machine-led decision-making, it’s a signal of something larger than a marketing trend, it’s the early shape of how purchasing itself is being restructured around AI systems rather than human browsing habits.
Three Different Shopping Models
The shift becomes clearer when you lay out how a purchase actually gets made under each model. Traditional ecommerce follows a familiar path: a human searches Google, lands on a website, evaluates a product, and checks out.
AI discovery changes the second step, a human asks an AI assistant a question, receives a recommendation, and follows that recommendation to a retailer before completing the purchase themselves.
Agentic commerce goes further still, a human delegates the task entirely, an AI agent pulls product and service data directly, compares options against stated criteria, and executes the transaction with minimal human input at each individual step.
According to Searchflex, an ecommerce search agency, that third model is where the real disruption sits, and it’s directly relevant to anyone working in robotics and automation, since it represents software agents making comparative decisions and taking action on a person’s behalf, the same underlying pattern found in autonomous systems more broadly.
The Data Behind the Shift
This isn’t confined to a single data point either. Adobe’s subsequent research covering the 2025 holiday shopping season found that AI-driven traffic to U.S. retail sites rose 693.4% year over year.
Visitors arriving through AI sources spent 45% longer on retail sites and viewed 13% more pages per visit than shoppers from other channels, while being 33% less likely to leave immediately compared to non-AI traffic.
That consistency across two separate measurement periods, seven months apart, is exactly the kind of pattern that separates a genuine structural shift from a short-lived spike tied to one product launch or news cycle.
The Real Visibility Question
As Searchflex puts it: “For ecommerce brands, the next visibility problem isn’t simply whether Google can crawl a product page. It’s whether AI systems can understand the product, differentiate it from alternatives, and confidently recommend it.”
That’s a materially different challenge to conventional SEO. A page can rank well in Google while still being poorly understood by an AI system attempting to compare it against competing products on a shopper’s behalf.
What Machines Actually Need to Understand a Product
Several categories of information likely feed into how confidently an AI system can represent and recommend a product, though it’s worth being upfront that nobody outside the AI labs themselves knows precisely which signals carry the most weight.
What’s reasonable to say is that these all form part of the information ecosystem machines draw on:
Structured product information, the kind that clearly defines what an item is, its specifications, and its variants, gives systems something unambiguous to work from.
Reviews and third-party brand mentions provide external validation beyond what a retailer says about its own product. Detailed specifications, accurate availability, and clear pricing reduce the chance of an AI system recommending something based on stale or incomplete data.
Merchant feeds and schema markup give machines a structured, machine-readable version of information that might otherwise be buried in unstructured page text.
Category relevance and broader brand authority round this out, helping a system understand not just what a product is, but where it sits within a wider market and how trustworthy the brand behind it appears to be.
None of this should be read as a checklist that guarantees AI visibility. It’s a description of the raw material these systems have available when forming a recommendation, and brands with more complete, structured, and externally validated information are simply giving these systems more to work with.
Where This Leads
If agentic commerce continues on its current trajectory, purchasing decisions increasingly won’t involve a human directly comparing product pages at all.
That has implications well beyond marketing, touching how products need to be described, structured, and validated for an audience that includes machines making decisions alongside people.
The growth data suggests this shift is already underway rather than theoretical, and brands treating it as a distant possibility risk being poorly represented in exactly the systems an increasing share of shoppers are starting to rely on.
Main image courtesy of Easy-Peasy.ai
