AI search referral traffic went from a rounding error to a line item
AI search is reshaping how shoppers move from question to purchase, and the numbers are moving fast.
Less than a year ago, retailers described AI referral traffic almost apologetically: under 2% of total sessions, a footnote. By the final months of 2026, that share had climbed to roughly 5% company-wide, with some retailers approaching 10%, according to Digital Commerce 360's interviews with four online retailers in September [2].
The shift is not isolated to a handful of sites. A January 2026 study from the IBM Institute for Business Value, run with the National Retail Federation across 18,000 consumers in 23 countries, found that 72% of shoppers still visit physical stores, while AI-assisted shopping is emerging: 41% use AI assistants to research products, 33% to look for reviews, and 31% to search for deals [3].
That is the context behind the move from SEO to AI search optimization, also called GEO, generative engine optimization: the discipline of making a catalog legible to a model that answers the question itself, rather than a crawler that merely indexes a page.
From indexing pages to settling the question
Traditional search engines were built to route traffic: index relevant pages, rank them, send the user to click. Generative engines do something categorically different, they try to resolve the user's question inside the conversation itself, without a click ever happening. That single design difference is what is compressing the top of the commerce funnel.
“They're looking for really in-depth product information that's well-structured. They need a lot more content than what typically meets the eye on a product description page.”
The practical effect is that generic, keyword-dense content built purely to rank is losing ground, because the model already has an answer before the user reaches a results page. EMARKETER's own retail analysts describe three distinct audiences retailers now have to write for at once: human shoppers who respond to visual storytelling, AI-assisted consumers who need clear factual product information a chatbot can quote, and autonomous AI agents.
Those agents require structured, machine-readable specifications, pricing and availability [1]. Each audience reads different content, and serving only the first one is what gets a brand omitted from the second and third.
The funnel didn't disappear, it compressed and rerouted
A shorter funnel is not a simpler one to measure. When AI introduces a shopper to a product, that shopper takes an average of 3.8 online steps before purchasing, compared with 1.6 steps for shoppers who never touched AI, according to IAB research cited by EMARKETER [1]. More steps, not fewer, because the verification habit AI triggers adds research rather than removing it.
That verification habit cuts against the fear that AI simply siphons traffic away from retailer sites. After interacting with AI during a purchase journey, consumers' visits to brand and marketplace websites increased 80%, per the same IAB data [1]. AI is functioning as a parallel discovery channel that drives people back to the retailer to confirm what the model told them, not a replacement for the retailer's own site.
The attribution problem is that almost none of this shows up cleanly in a standard media mix model built around last-click or first-click logic.
Trust, not technology, is the bottleneck
Consumers are using AI for shopping research in large numbers, but most of them do not yet take its word for it. Only 46% of consumers say they fully trust AI shopping recommendations, and 89% still take additional steps to verify what the AI told them before completing a purchase, according to IAB and Talk Shoppe research cited by EMARKETER [1].
Among shoppers who used AI anywhere in their journey, 95% took extra online steps to confirm the decision [1].
That trust gap is uneven across categories. Consumers accept AI guidance more readily for repeat, low-consideration purchases, groceries, replenishment, than for higher-stakes decisions like electronics or travel, where additional independent research remains the norm [1].
Where consumers actually turn to AI first
Shopping behavior in AI tools is concentrated at specific, researchable moments rather than spread evenly across the journey. The NRF-IBM study found that among consumers who use AI while shopping, 41% use AI assistants to research products, 33% to look for reviews, and 31% to search for deals [3].
Category matters too. EMARKETER's reporting on Boston Consulting Group data shows roughly 6 in 10 global consumers who have used generative AI prompted it for consumer electronics purchases, while about half did so for travel, apparel or entertainment [1]. Shopping-related genAI use overall grew 35% between February and November 2025, according to the same BCG figures [1].
The GEO opportunity is not evenly distributed: the categories with the highest AI usage are exactly the ones where product-data quality and third-party validation (reviews, press mentions, spec consistency) will decide who the AI recommends, and they deserve content and data investment first.
What retailers are already seeing in their own analytics
Four retailers that spoke with Digital Commerce 360 in September 2026 show the range of outcomes inside that average 5% figure, and the conversion story is not uniform.
| Retailer | AI share of traffic | What they're seeing |
|---|---|---|
| Bero | Under 5% | Growth slower than expected, "not exponential" [2] |
| Povison | About 5%, trending to 10% by early 2027 | AI-referred users research fast and expect accurate recommendations [2] |
| La Joya Jewelry | About 5-6%, up from 2-3% | Conversion over 13% from AI referrals vs. a 3% baseline [2] |
| Edible Brands | Under 10% | AOV $5.50 higher, but conversion (5%) trails the 10%+ site average [2] |
Does AI-referred traffic actually convert better than typical traffic?
It depends on the category and the intent behind the AI query, not on AI referral status alone. At a market level, AI-associated referral traffic to e-commerce sites rose 62% year over year through July 2026, and AI-referred shoppers generated 53% more revenue per visit than other shoppers, according to Adobe data reported by Digital Commerce 360 [2].
But that aggregate masks real variation, because AI-referred shoppers arrive having already done research, which does not automatically make them spend more or convert better than a site's existing high-intent visitors. La Joya Jewelry's founder attributes its lift to the trust transfer effect: a shopper arrives pre-sold because an AI already vouched for the retailer [2].
Edible Brands shows the opposite nuance: its core audience already arrives with unusually high purchase intent regardless of channel, so AI referrals do not stand out against its gift-driven baseline [2]. The lesson for any commerce team: benchmark AI referral performance against your own historical baseline, not against an industry average built from different categories.
What this means for commerce teams right now
Three actions matter more than a technical checklist. First, rebuild measurement to capture multi-step, AI-influenced paths instead of relying on last-click models that miss the rise in brand-site visits following an AI interaction [1]. Second, segment content deliberately for human shoppers, AI-assisted researchers, and machine agents, since each reads different signals before trusting a recommendation [1].
Third, treat physical retail and reviews as trust infrastructure, not legacy channels, since most consumers still shop in stores and AI's own recommendations lean heavily on review consensus and brand reputation to decide what to surface [3].
The retailers quoted here are not waiting for certainty before acting. Povison is already working to improve how often AI assistants recommend its furniture ahead of the holidays [2]. That is the practical shape of the GEO shift: not a single optimization task, but an ongoing adjustment to a journey that increasingly starts, and sometimes nearly finishes, inside a conversation the retailer never directly controls.
Sources
- [1]EMARKETER, FAQ on the path to purchase: What marketers need to know as AI and social commerce redefine buying — emarketer.com
- [2]Digital Commerce 360, How AI referrals are impacting traffic at 4 online retailers — digitalcommerce360.com
- [3]National Retail Federation, Own the agentic commerce experience — nrf.com
