Key Takeaways
- Riskified's Q1 2026 Agentic Commerce Pulse shows AI use for product discovery has spread to 61.5% of US/UK consumers, while discomfort with AI agents completing purchases on their behalf has climbed to 55%, a sharp reversal from the 70% who described themselves as comfortable in Q4 2025.
- Half of respondents (50.8%) say AI platforms should bear responsibility for unauthorized purchases, and 73.9% expect biometric or one-time password authentication every time, indicating consumer concerns have shifted from vague unease to concrete implementation requirements.
- While payment infrastructure responses such as AmEx's ACE developer kit and the J.P. Morgan x Mirakl partnership are advancing trust frameworks, ecommerce merchants must urgently implement granular permission delegation, mandatory authentication, and clearly stated liability terms.
The Numbers Make the Convenience-vs-Delegation Gap Impossible to Ignore
Riskified's Q1 2026 Agentic Commerce Pulse finds 61.5% of consumers use AI for product discovery but 55% are not comfortable with AI agents making purchases on their behalf.
www.digitaltransactions.netRiskified (NYSE: RSKD), a global leader in ecommerce fraud and risk intelligence, released the Q1 2026 edition of its quarterly Agentic Commerce Pulse on April 27, 2026. The survey of 2,000 consumers across the United States and the United Kingdom delivered a finding that complicates any tidy narrative of agentic commerce adoption: the curve is not simply trending up.
Discovery use of AI has reached 61.5%. Yet discomfort with AI agents completing transactions on a shopper's behalf now stands at 55%, a striking reversal from the 70% who said they were at least somewhat comfortable in Q4 2025.
Riskified Chief Marketing Officer Jeff Otto put it plainly in a statement: a widening gap between adoption and trust. Shoppers want the convenience and personalization AI delivers, but they are not yet willing to hand over control or accountability. That temperature gap surfaced sharply in just one quarter.
Reading the Q4 2025 to Q1 2026 Inversion
The most striking finding is how quickly consumer sentiment flipped.
| Metric | Q4 2025 | Q1 2026 | Shift |
|---|---|---|---|
| Comfort with AI agents purchasing on consumer's behalf | 70% at least somewhat comfortable | 55% not comfortable | Sharp reversal of trust |
| AI usage across the shopping journey | 73% used AI at some point | 61.5% used for discovery and recommendations | Use cases narrowing |
| Top concerns | Payment security 32%, privacy 26% (general unease) | Online fraud risk 53.9%, strong auth required 73.9% (specified) | Concerns becoming concrete requirements |
| Liability for unauthorized purchases | Not surveyed | AI platform 50.8%, retailer/brand 23.2%, self 18.7% | Platform-liability narrative dominates |
In Q4 2025, 70% expressed at least mild comfort with AI agents purchasing on their behalf. Three months later, 55% explicitly say they are uncomfortable. Same surveyor, same markets, same methodology, so it is reasonable to read this as accumulated lived experience rather than statistical noise.
The character of the concerns has changed too. Where Q4 cited "payment security 32%" and "privacy 26%" as diffuse worries, Q1 shows 53.9% specifically pointing to AI-driven fraud risk and 73.9% expecting biometric verification or one-time passwords every transaction. Vague unease has hardened into specific implementation requirements over a single quarter.
What the Six Headline Numbers Really Say
Each Q1 figure deserves a closer look.
| Number | Finding |
|---|---|
| 61.5% | Have used AI tools for product discovery and recommendations |
| 55.0% | Are not comfortable with AI agents making purchases on their behalf |
| 46.5% | Do not trust any company to manage purchases for them |
| 53.9% | Believe AI could increase the risk of online fraud |
| 73.9% | Expect biometric or one-time password authentication every transaction |
| 50.8% | Believe AI platforms should be responsible for unauthorized purchases |
The standout is 46.5% saying they do not trust any company to manage purchases for them. Nearly half of respondents reject the premise of delegation entirely, choosing neither AI platforms nor retailers nor banks. The dominant answer right now is "no one should be doing this on my behalf."
The liability breakdown is equally telling. AI platforms 50.8%, retailer or brand 23.2%, the consumer themselves only 18.7%. That distribution suggests shoppers frame agentic commerce less as "a purchase I actively made" and more as "a service I outsourced to AI." Responsibility, in their mental model, sits with the provider, not the buyer.
Why Trust Slid Backward in Just Three Months
Several forces likely produced the 15-point reversal.
First, real-world experience. Respondents who answered Q4 2025 with optimistic intent had a few months to actually try ChatGPT Instant Checkout or retailer-built AI agents, and to encounter mis-orders, unintended charges, and broken cancellation flows in production.
Second, the steady drumbeat of AI incident reporting. Prompt-injection attack scenarios against agentic shopping assistants have been published in detail by security vendors such as Palo Alto Networks Unit 42, raising baseline consumer wariness. As we covered in Agentic AI retail fraud risks, gift-card theft and automated return abuse are no longer theoretical, they are implementable threats.
Third, no consensus on where this should happen. Asked which platform they prefer for agentic commerce, 31.2% chose general AI tools such as ChatGPT or Gemini, 27.0% chose retailer sites or apps, and 24.4% answered "no" outright. One in four is opting out at the platform-selection stage, which is hard to dismiss.
Industry Response: AmEx ACE and J.P. Morgan x Mirakl
Closing the trust gap is, for now, mostly a payment-infrastructure project.
American Express in April 2026 introduced ACE, the Agentic Commerce Experiences developer kit. ACE registers AI agents in advance, restricts payment to verified agents using tokenized credentials, and pairs with a new Agent Purchase Protection program that gives cardholders a resolution path for erroneous registered-agent transactions. Because AmEx is simultaneously the issuer, network, and acquirer, the design is unusually vertically integrated.
J.P. Morgan Payments partnered with French fintech Mirakl SAS in March 2026 to embed payment processing, tokenization, and fraud protection into Mirakl's Nexus agentic commerce service. Prashant Sharma, J.P. Morgan Payments' executive director of biometrics and identity solutions, framed the goal as combining verified agent identity, user-controlled permissions, and bank-grade risk management so that AI agents can transact while trust remains intact.
Both moves share a common thesis: not "AI agents pay autonomously" but "every agent's identity and permission scope is explicit and auditable." Given that 73.9% of Riskified's respondents expect authentication every transaction, the registration-tokenization-protection trio aligns directly with stated consumer expectations.
What Merchants Should Build Now
The findings push merchants past the "should we offer agentic commerce" debate and into "how do we engineer trust." Several priorities stand out.
The most important is granular permission delegation. Rather than handing AI agents end-to-end purchase authority, merchants should expose finer controls: "agent can build the cart, human approves checkout," "full delegation only for subscriptions," and similar gradations. With 46.5% rejecting any delegation outright, an experience that assumes full delegation will repel rather than convert.
Second, per-transaction authentication as default. The 73.9% who demand biometrics or OTPs is not a wishlist, it is a de facto baseline. Agentic checkout flows should require Passkey, FIDO2, or OTP confirmation at the cart-lock step every time.
Third, explicit liability and purchase-limit pairings. Because 50.8% expect AI platforms to bear fraud liability, merchants running their own on-site agents need to spell out where merchant responsibility starts and ends, in their terms of service. The cleanest path is to mirror something equivalent to AmEx's Agent Purchase Protection through merchant terms or processor agreements.
As discussed in Trust and security frameworks for agentic commerce, agent identity verification, transaction limits, and audit logging are the minimum bar regardless of merchant size.
Closing Thoughts: The Trust Gap Is a Commercial Opportunity
The Q1 2026 Agentic Commerce Pulse makes one thing concrete. Adoption of agentic commerce will be paced by trust, not by technology. AI-driven discovery is genuinely spreading, but consumers have drawn a clear line between "add to cart" and "complete the purchase."
That line is also a competitive opening. Merchants that move first on granular delegation, default-on biometric authentication, and explicit liability framing can convert the trust gap itself into a moat.
Riskified's quarterly cadence makes this an unusually useful tracker for consumer psychology. Whether Q2 2026 shows trust recovering or rejection hardening will be a continuous signal worth watching for everyone building in agentic commerce.
