As we noticed on Black Friday, Amazon’s AI assistant, Rufus is reshaping shopper habits. Sessions that led to purchases doubled for Rufus customers, in contrast with solely a 20% enhance for consumers who didn’t use it.
That shift isn’t siloed to Rufus both. Seventy-three p.c of shoppers use AI assistants to assist them store, 58% use AI for vacation presents, and 70% are snug letting AI full their transactions, based on an October Riskified survey of 5,400 folks.
The way forward for AI buying is more likely to comply with one in all three paths, every with very totally different implications for retailers:
Merchant-led ecosystem: Some corporations will maintain AI buying native to their very own platforms, blocking exterior instruments. Amazon’s Rufus is the prototype of this method. This mannequin preserves management over pricing, buyer relationships, and model expertise whereas permitting fraud groups to keep up the behavioral indicators they depend on. The trade-off requires vital funding in AI capabilities to stay aggressive.
Collaborative ecosystem: Others will embrace partnerships with third-party AI buying instruments utilizing open protocols similar to ACP, AP2, and Visa VIC. Retailers retain some management over pricing and model presentation whereas benefiting from AI-driven site visitors. These frameworks additionally enable verification of each shopper and AI interactions, sharing intelligence to cut back fraud and disputes. Google’s AP2 announcement in September 2025, backed by Mastercard, American Express, and PayPal, indicators that the business is shifting on this course. Open requirements give retailers a seat on the desk and stop full disintermediation.
Decentralized buying: The most disruptive path is one wherein third-party instruments like ChatGPT develop into the first interface for shoppers. Retailers threat being diminished to undifferentiated stock suppliers competing on worth and achievement. Autonomous AI can result in elevated disputes, pleasant fraud, and chargebacks, significantly when accounts are compromised. A single hijacked profile can execute fraudulent purchases throughout a number of shops — a state of affairs we confer with as “Compromised AI-as-a-Service.”
This agentic evolution is remodeling the complete commerce funnel.
Even well-intentioned AI can miss transport or return insurance policies, producing non-fraud disputes. AI that scans a number of websites for the bottom worth can escalate pricing competitors. And, with fee safety topping shoppers’ considerations, retailers can’t afford to be passive.