The discovery layer has already flipped to AI
The data from RTB House, published by Retail Dive in mid-August, confirms something retail technology leaders have suspected but rarely seen quantified this cleanly. Sixty-eight percent of consumers used at least one AI platform for shopping-related tasks in the past three months, and 62 percent specifically used AI to compare prices, brands, and reviews. This is not an early-adopter curve anymore. It is a majority behavior, and it has happened faster than most retail marketing organizations have adjusted their budgets.
The more consequential number is the 59 percent of shoppers who say AI platforms helped them discover a brand they did not previously know. RTB House VP Jaysen Gillespie framed this as the end of an era where social feeds drove product discovery. For any retailer whose growth strategy still leans on Instagram and TikTok discovery spend, this survey is a signal to start testing AI-referral optimization now, before the channel matures and the cost of visibility rises the way paid social did a decade ago.
The transaction layer has not moved nearly as fast
Here is where the enthusiasm breaks down. Only 49 percent of consumers say they would let an AI agent complete a purchase on their behalf, and among people who do not already use AI tools regularly, trust in AI to handle a transaction collapses to just 3 percent. That is a near-total rejection from half the shopping population, far stronger than a soft preference for keeping a human in the loop. Any retailer building an agentic checkout flow on the assumption that discovery trust translates directly into transaction trust is building on the wrong data, and should expect adoption curves closer to this 49 percent ceiling than to the 68 percent who already use AI for research.
The generational split adds more nuance than most agentic commerce pitches acknowledge. Thirty-five percent of all respondents want a human review step before AI completes a purchase, and that rises to 44 percent among baby boomers, a demographic that still controls a disproportionate share of household spending. Meanwhile 42 percent of American millennials said they would let an AI agent make purchases up to 250 dollars if returns were guaranteed within seven days. That gap shows the trust threshold moves with price and policy rather than sitting fixed, and retailers should be building return-policy guarantees directly into their agent checkout UX rather than treating them as a separate customer service function bolted on after the sale.
AI Is Extending Shopping Decisions Rather Than Shortening Them
Forty-two percent of consumers say AI tools actually extend their decision-making time by surfacing more options to consider. That cuts against the pitch most agentic commerce vendors make, which is speed and friction removal. The honest read of this data is that AI is currently functioning primarily as a research amplifier for a large share of shoppers, widening the consideration set rather than compressing the path to purchase. Retailers optimizing checkout flows purely for speed may be solving a problem consumers do not currently have at the discovery stage, while leaving the actual bottleneck, an overwhelming volume of AI-surfaced options, completely unaddressed.
This matters for where technology investment should land in the next 12 months. If AI is extending research time, the highest-value retail AI investment right now is probably in making product data, comparison tools, and review aggregation better within your own properties, so that shoppers doing AI-assisted research land somewhere your own site can convert them. That is a more defensible use of budget than racing to deploy a one-click agent checkout that a majority of shoppers say they do not yet trust enough to use.
Trust in the AI platform itself is more fragmented than vendors admit
The survey found 43 percent of respondents trust Google AI Overviews and ChatGPT equally for shopping guidance, with Claude at 23 percent and Grok at 21 percent trailing well behind. That fragmentation matters operationally, because retailers building agent integrations have to decide which platforms to prioritize with limited engineering and merchandising bandwidth. A near-even split between the two leaders means betting exclusively on a single platform functions as a concentrated bet on one company's product roadmap, not the safe default choice it might look like on a slide deck to the board.
It also undercuts the idea that one dominant agent will control retail discovery the way Google dominated search. Fifty-nine percent of consumers still say they trust friends and family over AI tools for purchase decisions, ahead of the 44 percent who trust AI. Human recommendation remains the single most trusted channel, which should temper any retail strategy built entirely around AI-agent optimization at the expense of referral and community programs that already work.
What retail leaders should actually do with this data
The practical takeaway is sequencing. Retailers should be investing now in making their product data, pricing, and reviews maximally legible to AI research tools, because that behavior is already mainstream and growing across every demographic this survey measured. Agentic checkout deserves real investment too, but it should be built around the trust levers this survey identifies directly: guaranteed returns, spend caps tied to category and price point, and an easy human-review toggle that shoppers can turn on or off per order, rather than assuming frictionless autonomy is the default experience most shoppers actually want today.
The 1 trillion dollar agentic commerce projection for 2030 holds up fine against this data, but the timeline and shape of that growth curve look different than most vendor pitch decks assume. This data suggests the growth curve will be led by AI-assisted research converting at higher rates through traditional checkout, with fully autonomous agent purchasing scaling more slowly and unevenly by demographic and price point. Retailers who build for that staged reality, research-assisted conversion first and autonomous purchasing layered in gradually by segment, will convert better over the next two years than those betting the roadmap entirely on agents closing the sale from day one.
