Ask consumers whether they want AI to do their shopping and you get two answers from the same crowd. In a Checkout.com survey, 33% said they expect at least 10% of their purchases to be AI-driven within a year. In the same research, 24% said they will never hand a purchase to AI at all, and 27% said they trust no organization to run a shopping agent on their behalf.
Both groups are real. Both are sizable. And they are not the loud-versus-quiet split you might assume, where enthusiasts dominate the conversation and skeptics quietly come around later. These are firm, opposing positions held at the same moment in the same market.
Meanwhile the actual volume is tiny. The same survey puts AI agents in just 3% of transactions today. So we have a channel that is simultaneously overhyped by its boosters, written off by a quarter of buyers, and barely measurable in real sales. That is the adoption-trust gap, and it is the most useful thing to understand about agentic commerce right now.
The Readiness-vs-Reality Gap: 3% vs 89%
The single most jarring pair of numbers in the Checkout.com survey is this: AI agents touch 3% of transactions, and 89% of merchants are actively preparing for agentic commerce.
That is not a market. That is a bet. Nearly nine in ten merchants are pouring resources into a channel that, by their own industry's measurement, is rounding-error small today. The infrastructure exists. ChatGPT Instant Checkout, built on the Agentic Commerce Protocol from OpenAI and Stripe, launched in September 2025. Google's Universal Cart arrived in 2026. The rails are real and shipping. The traffic on them is not, yet.
Why pour money into 3%? Because merchants are not betting on today's volume. They are betting against being late. The same survey found 72% of merchants admit consumers will adopt faster than the industry is prepared. Read that carefully: the people building for agentic commerce mostly believe they are already behind. The 89% preparing and the 72% who feel under-prepared are largely the same anxious group.
- The infrastructure is mature, the demand is not. Protocols and checkout rails launched in 2025 and 2026. Transaction share is still 3%.
- Merchant urgency is driven by fear of timing, not current ROI. Preparing for a channel is cheap relative to missing it.
- "Preparing" covers a wide range. It can mean a full agentic checkout integration or simply cleaning up a product feed. The 89% figure flattens both.
The honest reading is that merchants are buying option value. That can be smart. It can also turn into expensive over-building if you mistake a hedge for a forecast.
The Consumer Trust Split: 33% vs 24% and 27%
On the buyer side, the gap is about trust, not awareness. People understand what an AI shopping agent is. A third are ready to use one. A quarter refuse, and more than a quarter trust nobody to operate one.
Who wants it
The 33% who expect at least 10% of their purchases to be AI-driven within a year are not fringe. That is a meaningful slice of the market signaling real intent. If even half of them follow through, the 3% transaction figure moves materially.
Who refuses
The 24% who say never and the 27% who trust no organization are the part the hype cycle keeps skipping. These are not people who have not heard of agentic commerce. They have heard of it and decided against it, at least for now. Their objection is delegation itself: handing money and judgment to software run by a company they did not choose to trust with that role.
The critical caveat: these are stated intentions, not measured behavior. People say they will never do things they later do, and say they will adopt things they never touch. The 33% optimists may not follow through. The 24% refusers may quietly change their minds when the experience is good enough. Treat all of these as directional signals about sentiment, not as a demand forecast you can build a P&L on.
Why Both Sides Can Be True at Once
It is tempting to call the survey contradictory. It is not. Several things are true together.
- Different buyers, different risk tolerance. The 33% and the 24% are not the same people changing their minds. They are different segments with genuinely different comfort levels around delegation and money.
- Category matters more than the average. A reorder of cat litter or printer ink is a low-stakes, repeat decision that suits an agent well. A first-time apparel purchase or a considered electronics buy is not. The same person can want AI for one and refuse it for the other.
- Trust is earned per use case, not granted up front. The 27% who trust no organization today are describing a starting position, not a permanent one. Trust in any new commerce channel tends to build transaction by transaction.
So the market is not confused. It is segmented and early. The mistake is averaging it into a single "consumers want AI shopping" or "consumers reject AI shopping" headline. Neither is accurate. We cover how AI systems actually surface and rank products, regardless of who is doing the buying, in how LLMs choose recommendations.
What This Means for Merchants Right Now
The gap creates a real planning problem. Over-invest in agentic checkout and you may build for traffic that takes years to arrive. Ignore it and you risk being invisible the moment the 33% start acting on their stated intent.
The way through is to separate two questions that often get bundled together. One is "should I bet on agentic checkout volume soon?" The answer there is: probably not heavily, given 3%. The other is "should I make sure AI systems can find, understand, and recommend my products?" The answer there is yes, today, because that work pays off in regular AI search long before agentic checkout matters. It is the same groundwork either way, which is why it is low-regret.
What Smart Merchants Do With This
The right moves are the ones that help you whether agentic commerce stays at 3% or jumps to 30%. Bet on the groundwork, not the timing.
- Get your product data AI-readable now. Clean titles, accurate attributes, correct pricing, real availability, complete structured data. Agents and AI search both consume this. If your feed is messy, you lose in both channels regardless of adoption speed. See our Shopify AI visibility complete guide for the specifics.
- Audit whether AI assistants can already find you. Before building anything new, check what ChatGPT, Gemini, and Perplexity say when asked to recommend products in your category. If you are absent or misrepresented, that is the cheapest, highest-leverage fix available.
- Watch the 3% as a signal, not a target. Track agentic transaction share in your own data if you can isolate it. Let real volume, not survey optimism, decide when to invest in deeper integration.
- Do not re-platform or re-architect for agents yet. The 89% who are "preparing" mostly are not rebuilding their stores. Match that. Light, reversible steps beat heavy, hard-to-undo ones while the channel is this small.
- Plan for the category split. If part of your catalog is low-stakes and repeatable, that is where agentic commerce lands first. Prioritize clean data and availability there before anything considered or high-ticket.
The throughline: every move above improves your standing in AI search that already drives traffic, so none of it is wasted even if agentic checkout stalls.
How AdsX Can Help
We work with Shopify and e-commerce brands on exactly this kind of low-regret positioning. We check whether AI assistants can find and correctly recommend your products today, fix the product data and structured markup that both AI search and future agents depend on, and help you size the agentic-commerce opportunity against the real 3% baseline instead of the hype. The goal is to make you discoverable now and ready later, without over-building for traffic that has not arrived. Run a free AI visibility audit to see whether AI assistants can find and recommend your products, or talk to our team about a low-regret agentic-commerce plan.
The agentic commerce gap is not a contradiction to resolve, it is a market to read accurately. A Checkout.com survey shows merchants preparing en masse for a channel that is still 3% of transactions, and consumers split between a third who expect AI shopping soon and a quarter who refuse it outright. Remember these are mostly stated intentions, not observed behavior. The winning move is the low-regret one: make your products findable and understandable by AI now, and let real volume, not survey optimism, decide when to build deeper.
Sources:
- Checkout.com: Consumer demand for AI shopping is forming fast, but trust for agentic commerce is still catching up
- Stripe and OpenAI: Instant Checkout (Agentic Commerce Protocol)
- Google: Universal Cart / shopping cart