TLDR;
Visa says more than half of US adults in its commissioned holiday survey have used AI for at least one shopping task, with the strongest comfort around comparing prices, checking availability and finding discounts. Payment is a different story: Visa separately reported that only 23% trust generative AI to transact for them. Brands should make product and offer data easy for assistants to compare, but keep checkout consent, fraud controls and human review obvious.
What happened
Visa's Holiday Spending Shift Survey was conducted online by Morning Consult among 1,016 US adults from September 26 to October 2, 2026, weighted to the adult population with a stated margin of error of three percentage points. Visa reported that 54% had used AI for at least one shopping task, up from 47% in its previous-year comparison. Comfort was highest for price comparison at 61%, availability checks at 59%, discount finding at 57% and gift ideas at 54%.
The provider release also points to a clear boundary between research and delegated action. Visa cited its separate Trust Index finding that only 23% of US consumers trusted generative AI to handle payment transactions. The survey measures what respondents say they have done or would accept. It does not show completed agentic purchases, conversion rates or incremental sales caused by an AI recommendation.
Independent retail analysis from EMARKETER highlighted the same discovery shift and argued that merchants need to be visible when consumers use AI to compare products and deals. A separate Morning Consult holiday report, based on different research, reached a similar directional conclusion: 53% had used an AI assistant for product search, comparison or purchase, but automatic purchasing comfort reached no higher than 31% across tested categories and only 25% were comfortable sharing payment details. Those figures are context, not an extension of Visa's sample.
Why it matters
The commercial signal is not that shoppers have handed the whole purchase to an agent. It is that the upper and middle parts of the buying journey are becoming machine-mediated first. A consumer can ask an assistant to compare three products, identify a local option and find a discount before ever visiting a merchant page. If the assistant sees stale inventory, vague variants or inconsistent prices, the brand can lose the shortlist before its usual analytics record a session.
That makes operational product truth more valuable than generic AI content. Assistants need stable identifiers, current price and stock, variant relationships, shipping constraints, promotion dates, return terms and evidence for product claims. Enterprise teams should expect the same product to be represented through their own site, retailer feeds, marketplaces and AI platform integrations. Conflicts between those systems can create a poor recommendation even when each channel looks acceptable in isolation.
The trust gap also affects checkout design. An assistant may be useful for research while the shopper still wants to confirm the seller, total cost, delivery address and payment method. Hiding those controls to make an agentic flow look seamless can increase abandonment, disputes and fraud risk. The better design is progressive delegation: automate low-risk comparison and cart building, then make authorization and final commercial terms unmistakable.
How your brand can benefit / be affected
Start with a shopping-data audit built around the tasks in the survey. Test whether an assistant can distinguish price from promotional price, identify in-stock variants, explain discount eligibility and return the right local availability. Compare the answer with the product page and checkout. Record where the assistant sourced each fact and which system owns the correction when the answer is wrong.
Next, separate AI-assisted discovery from agent-completed payment in measurement. Add consented referral and checkout fields that identify an assistant or agent where possible, but do not infer a completed agentic purchase from a conversational referral alone. Track shortlist visibility, product-page arrival, cart creation, checkout start, human confirmation, authorization outcome, fraud review and post-purchase disputes as distinct stages.
Finally, keep the evidence in proportion. Visa's results are useful US survey data commissioned by a payments provider, not transaction telemetry or a global forecast. Use them to prioritize product-data quality and controlled checkout experiments, then validate behaviour with your own logs, customer research and payment outcomes. The brands that benefit will be the ones assistants can compare accurately and shoppers can still trust when money moves.
News date: 30 October 2026. Editorial review: 2 November 2026.