Milena Traikovich has spent the last several years at the intersection of consumer psychology and digital performance, carving out a reputation as one of the most forward-thinking voices in demand generation. As businesses grapple with a landscape where traditional search is no longer the sole gatekeeper of commerce, her expertise in navigating the “black box” of AI-mediated discovery has become indispensable. Today, she sits down with us to discuss the seismic shifts in how shoppers interact with generative AI platforms like ChatGPT and Gemini, and how a new measurement framework is finally shedding light on the hidden path from an AI chat to a completed purchase.
We are seeing a massive shift where consumers research products within AI assistants, but the data often goes dark before they hit a retailer’s site. How do you account for this “dark traffic” when trying to measure the true impact of AI on sales?
Accounting for this “dark traffic” requires moving beyond the traditional reliance on direct referral links, which are increasingly becoming an incomplete metric for success. Recent data reveals a startling gap: while American consumer visits to AI platforms surged by 28.6% between January 2025 and January 2026, the direct referral traffic from these services remained almost entirely flat. This tells us that shoppers are treating AI as a deep research hub rather than a simple click-through portal. When we look at the research, about 55.9% of the visits that follow an AI brand recommendation actually arrive through a traditional search engine later on, rather than a direct link from the chat. To bridge this gap, we are looking toward the new measurement system being launched in the fourth quarter of 2026 by NIQ and Similarweb, which is designed specifically to connect those early-stage AI conversations with subsequent website visits. By tracking the specific products recommended within ChatGPT or Gemini and then matching that data with retail sales and consumer behavior patterns, we can finally see the invisible thread that links a query about “best hiking boots” to a purchase made three days later via a direct search.
The data suggests that shoppers arriving from AI sources are significantly more valuable than those from traditional channels. What do you think explains this massive spike in revenue per visit from AI-referred consumers?
The numbers are quite staggering, with shoppers referred through services like ChatGPT and Gemini generating 53% more revenue per visit than those coming from non-AI sources. I believe this boils down to the high level of intent and the “pre-vetted” nature of these visitors. By the time a consumer clicks a link or searches for a brand after an AI interaction, they have already bypassed the broad discovery phase; they’ve likely compared features, checked reviews, and filtered by price within the AI interface. We see that 34% of consumers are using these assistants specifically for product research before they ever go looking for a deal. This means that when they finally land on a retailer’s site, the psychological friction of the purchase has already been smoothed over. They aren’t just “browsing” anymore; they are arriving with a high-conviction mindset, having been guided by a personalized recommendation that feels more like a concierge service than a traditional advertisement.
With the introduction of tools like OpenAI’s Instant Checkout and Google’s Universal Cart, the shopping journey is moving entirely inside the AI platform. How should brands adapt their digital infrastructure to stay competitive in this “agentic” commerce environment?
Brands can no longer afford to view their website as the only place where a transaction happens; they must prepare for a future where the AI agent is the primary customer. The rollout of OpenAI’s Instant Checkout in 2025 was a major signal, using the Agentic Commerce Protocol to let users buy products without ever leaving the chat. Google is following suit this summer in 2026 by bringing the Universal Cart to the Gemini app and Search, allowing for a seamless checkout experience across their entire ecosystem. To stay competitive, brands must ensure their product information is “AI-ready”—meaning it is structured and complete so that these assistants can accurately identify and represent their inventory. This involves more than just SEO; it’s about participating in open standards like the Universal Commerce Protocol so that real-time pricing, inventory, and cart functions can be accessed by AI agents. If your backend isn’t structured to talk to these agents, you will simply disappear from the recommendations, regardless of how good your product is.
It is fascinating that nearly 75% of shoppers are now using AI during their discovery phase. Given this level of adoption, what specific metrics should marketing teams be looking at to determine if their brand is actually winning the AI recommendation game?
Traditional metrics like “impressions” are becoming obsolete in a world where an AI might mention your brand in a private conversation with a user. Instead, we have to look at “visibility in recommendations” and “share of voice” within the AI response itself. The new NIQ and Similarweb framework will be essential here, as it measures exactly which products appear in AI suggestions and how those brands compare to their competitors in real-time. We also need to track the “latent conversion” rate—specifically, the 2.5 times higher likelihood of a site visit occurring within seven days of an AI recommendation. Marketing teams should be analyzing the gap between their AI mentions and their direct/organic search traffic to see if they are successfully capturing the 55.9% of users who research in AI but don’t click a direct link. If 20% of consumers are already using AI as a core part of their shopping process, the most critical metric is no longer the click-through rate, but the “content readiness score”—a measure of how effectively an AI can parse and recommend your product based on the structured data you provide.
What is your forecast for the evolution of AI-assisted shopping as we head into the final months of 2026?
I expect we are going to see the complete collapse of the traditional “marketing funnel” in favor of a more circular, instantaneous shopping experience. By the end of 2026, the distinction between “searching” and “buying” will have blurred to the point of being unrecognizable for a large segment of the population. As the NIQ and Similarweb measurement system rolls out to more categories and markets, we will finally have the data to prove that AI isn’t just a research tool—it is the new storefront. We will likely see a massive shift in budget allocation, where brands move money away from traditional search ads and into “AI Optimization” to ensure they are the first choice when a user asks Gemini to “find the best eco-friendly coffee maker and buy it for me.” The brands that win will be those that embrace this “agentic” world, making their products easily discoverable, comparable, and purchasable within a single conversational interface. The era of clicking through ten different tabs to compare prices is ending; the era of the AI concierge is officially here.
