Retailers have spent the past two years preparing for a future where AI agents handle the shopping. But consumers have other plans. They’re happy to let AI do the legwork, but much less willing to let it make the final call.
That’s the key takeaway from new Circana research, presented at the company’s recent “AI Measured” webinar on where consumers draw the line with the technology. Based on a September survey of more than 4,600 U.S. adults, the study suggests shoppers are neither fully embracing AI nor rejecting it. Instead, they’re making practical trade-offs, one step at a time.
“Consumers will hand AI the map to their next purchase, but not the wheel,” said Kiara Barrett, who leads global thought leadership at Circana.
The numbers support that. Fifty-three percent of consumers are comfortable with AI comparing prices, and 46% are comfortable with AI suggesting products. Comfort drops to 30% when AI chooses the item and just 16% when it completes a purchase without approval. Only 1% to 2% say AI already places orders for them.
That gap — between helping and deciding — should guide how retailers approach the agentic future. As Barrett argued, autonomy “will be earned task by task, not adopted as one giant leap.”
In the meantime, AI is delivering real value. About a third of shoppers say it has strongly influenced how much time and money they save and how confident they feel in their choices. Among shoppers who use AI, 60% find it extremely or very useful. What AI isn’t doing, at least not yet, is increasing cart size: Only 24% say it has strongly influenced the size of their basket.
Sally Lyons Wyatt, Circana’s global consumer goods and food service chief advisor, summed it up: “Consumers are ready for a copilot that simplifies decisions, not an autopilot that removes them.”
Trust, not technology, is the bottleneck. Thirty-four percent of shoppers say they always or often check information from AI shopping tools against other sources. Retailers shouldn’t assume their brand equity carries over to their own assistants: Just 22% of consumers trust AI assistants on retailer sites, seven percentage points fewer than trust general-purpose AI tools. Shoppers may suspect these assistants prioritize what the store wants to sell over what’s best for them.
Some of that distrust is self-inflicted. In about a quarter of the negative experiences Circana reviewed, the problem was outdated pricing, an expired sale or unavailable inventory. “Many AI trust problems are actually product data, pricing and inventory problems wearing the AI costume,” Barrett said.
The way forward is less about moving fast and more about getting the basics right. That means improving product data, explaining why items are recommended, labeling sponsored placements, and making automation opt-in and reversible. Predictable, low-stakes repeat purchases — such as cleaning products and household supplies — are a sensible place to start.
The retailers that come out ahead in this phase won’t be the first to hand AI the wheel. They’ll be the ones who give shoppers a good reason to trust AI with the map.