The Agentic Commerce Trust Gap: Only 23% of Shoppers Trust AI to Pay (2026 Data)
Shoppers welcome AI that helps them decide, but few will let it pay: 23% trust GenAI with payments, and only 5% want an agent buying on its own. Caps, returns and instant revocation change that. Here's the 2026 survey data from Visa, Checkout.com, Cover Genius and RTB House, and a spend mandate you can enforce in code.
- Published
- Reading time
- 9 min read

On this page
TL;DR: Shoppers like AI that helps them decide, but few will let it pay. Only 23% of US consumers trust GenAI to handle payments for them (Visa, September 2026). Only 5% are comfortable with an agent buying without asking (Cover Genius, September 2026), and 24% say they'll never delegate a purchase to AI (Checkout.com, June 2026). Trust rises sharply when there are safeguards: 42% of US millennials would let an agent buy within a $250 budget if they could return the item within seven days, compared with about a third without that return option (RTB House, August 2026). Consumers' own top requirements are spending caps (30%), instant revocation (29%) and easy cancellation (28%). The takeaway for anyone building agentic commerce: don't ask shoppers for trust up front. Let the agent earn it with limits that are enforced in code. Below: the data, what it means, and a working spend-mandate implementation.

This is one of four deep dives that follow my overview, Agentic AI in October 2026: Agents Now Need Owners. The others cover OpenAI Dots, Atlassian AMP and Oracle Fusion Claw.
The Trust Gap in One Table
Jordan McKee's Forbes column put a name to it: agentic commerce has a consumer trust problem [5]. Shoppers want AI across the shopping journey, but far fewer will let an agent complete the transaction. The 2026 surveys agree:
| Finding | Source |
|---|---|
| 23% of US consumers trust GenAI to handle payment transactions for them | Visa Trust Index, Harris Poll, 2,065 US adults, May 2026 [1] |
| 78% are open to an AI agent helping them shop, but only 5% are comfortable with it buying unilaterally | Cover Genius / Gather, 1,392 consumers, 7 countries, Sept 2026 [2] |
| 27% trust no organization to run an AI shopping agent, and 24% will never delegate purchases to AI | Checkout.com, June 2026 [3] |
| 35% want a human to review transactions before an agent buys (44% of baby boomers) | RTB House, 1,800+ consumers in US, UK, Japan and France, Aug 2026 [4] |
| 59% trust friends' and family's opinions when buying, vs 44% who trust AI tools | RTB House [4] |
| Almost two-thirds of retail leaders don't expect shoppers to fully accept agents buying for them before 2028 | Deloitte 2026 Retail Industry Global Outlook, via Cover Genius [2] |
Surveys word their questions differently, so the exact percentages don't compare directly. They all point the same way, though: help is welcome, but autonomy isn't, yet.
Trust Is a Ladder, Not a Switch
The Cover Genius breakdown is the most useful number set, because it shows how much autonomy people want. Of the 78% who are open to AI shopping help [2]:
| Level of autonomy | Share of consumers |
|---|---|
| Agent only recommends | 26% |
| Agent proposes and waits for approval | 32% |
| Agent buys within rules I set | 15% |
| Agent buys on its own | 5% |
That's a ladder. Most shoppers are on the first two rungs today. The product design job is to make the next rung feel safe enough to step onto.
The Safeguards That Move People
The most encouraging data shows that specific safeguards change behavior:
- Returns: 42% of US millennials would let an agent buy for them within a $250 budget if they could return the purchase within seven days. Without the return safeguard, it drops to about a third [4].
- Caps, revocation, cancellation: consumers' top non-negotiables are spending caps (30%), instant revocation (29%) and easy cancellation (28%) [3]. 75% of merchants agree that real-time revocation will be critical [3].
- Per-purchase limits: on average, consumers would let an agent spend £177 per purchase without extra approval, close to the £200 merchants assume [3].
- Brand loyalty is weaker than you'd think: 57% would let an agent switch brands for a better-value option [3].
- Recourse: in Cover Genius's survey, automatic payouts when something goes wrong were the biggest motivator, and the company argues a shopper's real protection is recourse, not their card [2].
- Category matters: people would delegate groceries (41%) and household supplies (31%) far sooner than financial services (15%) [3].
Where the Market Actually Is
Agents are still a small share of real transactions. UK and US merchants report that 3% of transactions involve AI agents, even though 89% of merchants are preparing for agentic commerce [3]. Merchants also feel behind: 72% say consumers will adopt agent-led shopping faster than most merchants are ready for [3].
Discovery has moved faster than checkout. In the US, Google AI Overviews and ChatGPT are the top AI shopping tools (43% each), followed by Claude (23%) and Grok (21%) [4]. McKinsey and ICSC project US agentic commerce could reach $1 trillion by 2030 [4]. At the same time, 42% of US consumers say AI tools extend the time it takes them to decide [4]. AI is changing how people research products well before it changes how they pay.
On the payments side, the major networks and platforms have been building agent infrastructure since 2025: Visa Intelligent Commerce, Mastercard Agent Pay, Google's Agent Payments Protocol (AP2), and OpenAI and Stripe's Agentic Commerce Protocol [6]. These standards address the same thing consumers are asking for: proof that an agent is authorized, for how much, and for what.
And brand trust carries over. 61% of US consumers say they'd trust Visa to handle agentic transactions, rising to 71% among frequent AI users [1]. Shoppers don't trust "AI" in general, but they do trust specific, accountable brands.
What This Looks Like From Fashion Commerce
I build generative-AI systems for fashion commerce at Modelia, and the data matches what I see. Fashion is close to the worst case for autonomous checkout: fit and taste are personal, returns are common, and a wrong size ruins the whole experience.
That's why the most valuable AI in fashion today helps people decide: showing a garment on a model with a similar body, answering fit questions, and building outfits. That's rungs one and two of the ladder. Autonomous buying makes sense first for replenishment, such as the same basics in the same size, where the risk is low and a return window covers mistakes.
Build It: A Spend Mandate Enforced in Code
Every safeguard consumers ask for (caps, categories, return windows, expiry and one-tap revocation) can be enforced deterministically at checkout instead of promised in a prompt. Here's a minimal mandate check. It runs with npx tsx spend-mandate.ts on Node 22+:
// spend-mandate.ts: run with `npx tsx spend-mandate.ts` (Node 22+)
// A shopper's mandate for an AI agent: the controls consumers say they need,
// enforced at checkout instead of promised in a prompt.
interface Mandate {
shopper: string;
perPurchaseCapUsd: number;
monthlyCapUsd: number;
allowedCategories: string[];
requireReturnWindowDays: number;
expiresAt: Date;
revoked: boolean;
}
interface Purchase {
merchant: string;
category: string;
amountUsd: number;
returnWindowDays: number;
}
type Verdict = { ok: true } | { ok: false; reason: string; askShopper: boolean };
function check(m: Mandate, p: Purchase, spentThisMonth: number, now = new Date()): Verdict {
if (m.revoked) return { ok: false, reason: "mandate revoked by shopper", askShopper: false };
if (now > m.expiresAt) return { ok: false, reason: "mandate expired", askShopper: true };
if (!m.allowedCategories.includes(p.category)) {
return { ok: false, reason: `category "${p.category}" not allowed`, askShopper: true };
}
if (p.returnWindowDays < m.requireReturnWindowDays) {
return { ok: false, reason: `only ${p.returnWindowDays}-day returns`, askShopper: true };
}
if (p.amountUsd > m.perPurchaseCapUsd) {
return { ok: false, reason: `${p.amountUsd} is over the ${m.perPurchaseCapUsd} per-purchase cap`, askShopper: true };
}
if (spentThisMonth + p.amountUsd > m.monthlyCapUsd) {
return { ok: false, reason: `would pass the ${m.monthlyCapUsd} monthly cap`, askShopper: true };
}
return { ok: true };
}
const mandate: Mandate = {
shopper: "user_42",
perPurchaseCapUsd: 150,
monthlyCapUsd: 300,
allowedCategories: ["groceries", "household", "basics"],
requireReturnWindowDays: 7,
expiresAt: new Date("2026-12-31"),
revoked: false,
};
const basket: Purchase[] = [
{ merchant: "FreshCart", category: "groceries", amountUsd: 86, returnWindowDays: 7 },
{ merchant: "HomeGoods", category: "household", amountUsd: 140, returnWindowDays: 30 },
{ merchant: "StyleHub", category: "fashion", amountUsd: 60, returnWindowDays: 30 },
{ merchant: "BasicsCo", category: "basics", amountUsd: 95, returnWindowDays: 14 },
{ merchant: "FlashDeals", category: "household", amountUsd: 20, returnWindowDays: 0 },
];
let spent = 0;
for (const p of basket) {
const v = check(mandate, p, spent, new Date("2026-10-11"));
if (v.ok) spent += p.amountUsd;
console.log(`${p.merchant.padEnd(10)} ${String(p.amountUsd).padStart(3)} ${v.ok ? "BUY" : `${v.askShopper ? "ASK" : "STOP"}: ${v.reason}`}`);
}
console.log(`spent this month: ${spent}`);
mandate.revoked = true; // one tap in the app
const after = check(mandate, basket[0], spent, new Date("2026-10-11"));
console.log(after.ok ? "BUY" : `after revoke: STOP: ${after.reason}`);Output:
FreshCart $ 86 BUY
HomeGoods $140 BUY
StyleHub $ 60 ASK: category "fashion" not allowed
BasicsCo $ 95 ASK: would pass the $300 monthly cap
FlashDeals $ 20 ASK: only 0-day returns
spent this month: $226
after revoke: STOP: mandate revoked by shopperHow it maps to the research:
- Per-purchase and monthly caps: the top consumer non-negotiable [3]. BasicsCo is blocked because it would pass the monthly cap, even though it's under the per-purchase cap.
- Return window: the safeguard that moves millennials from about a third to 42% [4]. FlashDeals has no returns, so the agent asks first.
- Category allowlist: start with groceries and household, where shoppers are most willing to delegate [3]. Fashion here goes back to the shopper, which is the right call for fit-sensitive purchases.
- Revocation: a single flag that stops everything immediately, which 29% of consumers and 75% of merchants say they need [3].
- "ASK" instead of "fail": a blocked purchase becomes a one-tap approval request, so the agent moves the shopper up the ladder rather than leaving them stuck.
In production, store the mandate on your server (never in the agent's context), check it in the payment path, and log each verdict. My agentic AI overview has a broader version of this pattern for any tool call, not just payments.
My Take
The trust gap isn't a problem marketing can fix. It's a product design constraint, and the data says how to design for it. Start at "recommend," move to "propose and approve," then allow "buy within limits," with caps, returns and revocation built in at every step. Each step needs evidence that the previous one worked.
The companies that win agentic commerce won't be the ones with the most autonomous agent. They'll be the ones whose agent shoppers trust with $50 this month and $150 next month, because it kept to its limits.
If you're building commerce agents on Shopify, my post on AI self-review for Shopify App Store submissions covers the review side, and the Shopify app review skill covers data-handling rules your agent will need to follow.
References
- 01Visa, "Visa Trust Index explores agentic commerce adoption," September 9, 2026 (Harris Poll, 2,065 US adults, fielded May 26–28, 2026). corporate.visa.com
- 02Cover Genius, "Do Consumers Trust AI Shopping Agents With Checkout?", September 25, 2026 (with Gather; 1,392 consumers, 7 countries). covergenius.com
- 03Checkout.com, "Consumer demand for AI shopping is forming fast but trust for agentic commerce is still catching up," June 9, 2026. checkout.com
- 04Retail Dive, "Consumers warm up to agentic AI purchases," August 13, 2026 (RTB House survey and McKinsey/ICSC report). retaildive.com
- 05Jordan McKee, "Agentic Commerce Has A Consumer Trust Problem," Forbes, October 1, 2026. forbes.com
- 06Eco, "Mastercard Agent Pay vs Visa Trusted Agent 2026: Compared." eco.com
Harsh Rastogi is an AI Product Engineer at Modelia, building production generative-AI systems for fashion commerce, and the creator of carcode. He writes about AI systems, developer tooling and production engineering at harshrastogi.tech.
Frequently asked questions
Do consumers trust AI agents to make purchases?
Mostly not yet. The September 2026 Visa Trust Index found only 23% of US consumers trust GenAI to handle payments for them, Cover Genius found only 5% are comfortable with an agent buying unilaterally, and Checkout.com found 24% say they will never delegate purchases to AI.
What makes shoppers trust AI shopping agents more?
Specific safeguards. Consumers' top non-negotiables are spending caps (30%), instant revocation (29%) and easy cancellation (28%), per Checkout.com. RTB House found 42% of US millennials would let an agent buy within a $250 budget with a seven-day return window, against about a third without it.
How much would consumers let an AI agent spend?
Checkout.com found consumers would, on average, let an agent spend about £177 per purchase without additional approval, close to the £200 merchants assume.
What share of transactions involve AI agents today?
UK and US merchants surveyed by Checkout.com report that about 3% of transactions involve AI agents, even though 89% of merchants are preparing for agentic commerce.
Which purchases will people delegate to AI first?
Low-risk, repeat purchases. Checkout.com found 41% would delegate grocery shopping and 31% household supplies, compared with 15% for financial services.
How should I build an AI shopping agent that people trust?
Move up an autonomy ladder: recommend, then propose and wait for approval, then buy within rules the shopper sets. Enforce per-purchase and monthly caps, category allowlists, return-window requirements and one-tap revocation in code at checkout, not in the agent's prompt.
- Agentic Commerce
- AI Agents
- E-commerce
- Agentic AI
- Payments
- Consumer Trust
Written by Harsh Rastogi, AI Product Engineer and Business AI Head at Modelia. More on AI products, agents and production engineering on LinkedIn.