What happened
BigCommerce and PayPal surveyed about 3,000 consumers across the US, UK, and Australia on agentic shopping. Around two-thirds said they're interested in AI shopping tools — a mainstream number, not an early-adopter fringe. But the interest came with a sharp shape: shoppers want AI handling product discovery — finding options, comparing them — and they're far cooler on autonomous purchasing, where the AI completes the transaction on its own.
The trust findings match that split. Fraud protection and purchase confirmation topped the list of what shoppers require before letting an agent near a transaction. Translated into plain terms: take the research off my plate, but I press the buy button, and I want a safety net if something goes wrong. That's a research assistant they're describing, not a proxy holding their card.
Why this matters
The industry narrative says agents will shop for people end to end. The people just said otherwise. That gap should decide where your next dollar of AI-readiness work goes: the comparison layer pays now; agentic checkout plumbing pays later, if ever. Two-thirds interest in AI-assisted discovery means the AI research session already exists somewhere in your funnel today. The autonomous purchase mostly doesn't, and shoppers are telling you they don't want it yet.
Winning discovery is concrete work, not strategy-deck work. If an AI is doing the comparing, being comparable is the ranking factor: spec tables that exist as real HTML, honest differentiators stated in words rather than implied by photography, plain answers to the questions buyers actually ask before choosing. Stores that make the comparison easy get recommended. Stores that make it hard get summarized from a competitor's page — and the competitor writes the framing.
The confirmation preference is quietly good news for merchants. If shoppers insist on approving the purchase, then your brand, your product page, and your checkout still exist in the loop — full autonomy would have commoditized all three into an API call. One honest caveat: surveys measure stated comfort, and stated comfort drifts. Cards online, mobile pay, one-click ordering — every trust threshold started as something most people swore they'd never do. Build for the preference shoppers have now; don't assume it's permanent.
What to do about it
Build the comparison layer on your money pages
Put spec tables in HTML on your top product pages and state the honest difference against the obvious alternative. Write the answer to "which one should I get" somewhere an AI can quote it verbatim — that's the exact question discovery-mode shoppers are handing to their assistants.
Make trust signals machine-readable
Shoppers ranked fraud protection and purchase confirmation as prerequisites, so put your guarantees where machines can read them: returns policy as crawlable on-page text plus MerchantReturnPolicy schema, warranty terms on the page rather than buried in a PDF. If an AI can't verify your safety net, it can't relay it.
Park the agentic checkout project
Autonomous purchasing is the weak half of this data, so treat checkout-agent integrations as a watch item, not a build item. Revisit when a surface with real transaction volume demands it. Until then, the same budget does measurably more in the comparison layer.