Zeta Global released new agentic commerce research on June 30, 2026, finding that 70% of AI shoppers still prefer to complete purchases on a brand’s own website rather than buy through AI.
The study surveyed 2,000 U.S. adults who had used AI to make a purchase within the previous three months. The sharper finding for marketers is that 54% of AI shoppers would choose a brand’s personalized AI experience over a general-purpose AI tool, rising to 58% among consumers ages 18 to 45.
For ecommerce and B2B commerce teams, the point is not that AI has replaced the storefront. The point is that AI is increasingly shaping discovery before the storefront visit, while the brand-owned experience still has a chance to win the purchase and the relationship.
Key Takeaways
- Zeta found that 70% of AI shoppers prefer buying on a brand website rather than through AI.
- Fifty-four percent would choose a brand’s personalized AI experience over a general-purpose AI tool.
- Parents are early adopters: 43% would let AI make purchases within a set budget.
- The survey covered 2,000 U.S. adults who used AI for a purchase in the previous three months.
- The B2B commerce read: measure AI-assisted discovery separately from final checkout.
What Zeta Found About AI Shoppers
Zeta’s research says agentic commerce is moving fastest in the research and decision stage. Thirty-six percent of AI shoppers now spend less time researching purchases, while 59% say AI has reduced the likelihood that they will return a purchase.
That does not mean shoppers are ready to hand over the full transaction. Zeta says only 21% spend more money because of AI, and the 70% brand-site preference shows that the checkout moment still carries trust, control, and habit.
The parent segment is the strongest signal. Zeta found that 43% of parents with children under 18 would allow AI to buy within a set budget, compared with 27% of non-parents. The same 43% would let AI automatically reorder household essentials, compared with 31% of non-parents.
AI Is Becoming the Decision Layer, Not Just Checkout
The hidden catch is that brands may keep the transaction while losing control of the decision that leads to it. If an AI assistant narrows the product set before the customer opens a website, the brand site becomes the confirmation layer, not the discovery layer.
Braze’s agentic commerce guide frames the same issue around machine-readable product data, structured attributes, and the brand relationship that survives AI intermediation. Axios’ Cannes coverage made the urgency more blunt: brands were told to act fast before AI agents become the first customer interface.
That is why Shopware’s merchant-control architecture matters for this story. Zeta is showing the demand side: consumers want AI help, but many still want the brand to own the final experience. Shopware showed one supply-side answer: keep catalog, rules, actions, payments, and permissions inside a joined commerce context.
The Trust Gap Has Not Gone Away
Zeta’s numbers should be read beside the trust research IVRIS has already covered. In Horizon Media’s AI shopping study, 68% of respondents distrusted an AI agent when a purchase lacked human input. That makes the 70% brand-site preference less surprising.
Our read: brand-owned AI is not a vanity feature. If shoppers trust the brand more than a general-purpose assistant, the brand’s own AI experience becomes a retention surface, a data surface, and a proof surface.
There is a second risk. Skyword’s AI brand-information research found that only 29% of consumers trust a brand outright when its message conflicts with what an AI tool says. A brand-owned AI experience helps only if it matches the evidence customers can verify elsewhere.
TechRadar’s B2B procurement argument adds another useful angle: the more complex the purchase, the more important structured data and decision rules become. That is where B2B commerce teams should be paying attention.
What B2B Commerce Teams Should Do Now
- Separate assisted discovery from final purchase. Track prompts, AI referrals, direct brand-site sessions, and checkout conversion as one connected journey.
- Audit product data for machine use. Agents need attributes, pricing logic, inventory status, return rules, and compatibility data they can read without guessing.
- Build a brand-owned AI test. Do not start with full delegated buying. Start with product guidance, reorder help, compatibility checks, and quote preparation.
- Measure trust after the transaction. A lower-return rate is useful, but so is whether the customer felt informed, in control, and willing to buy again.
- Keep AI visibility reporting tied to search reporting. Goodwater’s AI-first search finding shows why the first answer can shape the shortlist before a click happens.
Zeta’s research makes the near-term playbook clearer. Brands do not need to choose between AI commerce and owned commerce. They need to connect the two before assistants make the brand-site visit a late-stage formality.
