Key Takeaways
- Walmart reported that conversions on ChatGPT-based Instant Checkout were three times worse than on its own website, putting a number on the "trust gap" facing early agentic commerce
- Stripe is rolling out "Shared Payment Tokens" for agents, while Hopper Technology Solutions separates read from write actions — both signal that the technical guardrails are being built in earnest
- The PhocusWire experts agree that "behavioural design matters more than technology" — presenting the booking option at the right moment of readiness is what will determine winners
Walmart: ChatGPT Conversions Three Times Worse Than Own Site
Payments architecture is evolving as AI giants and travel brands seek to enable AI-powered booking, but a trust gap remains.
www.phocuswire.comTravel-industry publication PhocusWire published an analytical piece on the "trust gap" in agentic commerce on April 9, 2026. The trigger was Walmart's report that conversions on ChatGPT-based Instant Checkout came in at one-third the rate of its own website.
The number arrived shortly after OpenAI relocated its Instant Checkout inside ChatGPT, and signals that consumers need time to adapt to a new commerce surface. PhocusWire frames the result calmly: "agentic commerce isn't dead, but it isn't going to become a standard transaction method overnight either."
"Trust Doesn't Transfer Automatically Through a New Interface"
Speaking to PhocusWire, Matthew Mamet — a fractional chief product officer and chief growth officer — said: "Trust doesn't transfer automatically through a new interface. You have to earn it in the new context, which takes time and usually means worse numbers before better ones."
Mamet drew on his time leading the Instant Booking effort at Tripadvisor in 2015. The security architecture was state of the art back then, he recalled, but changing consumer habits is a different problem. "When a customer lands on Walmart.com, they're in Walmart's environment. The reviews are Walmart's reviews. The return policy is Walmart's. The whole context signals: this transaction is safe." Inside ChatGPT, that surrounding context melts away and the customer is purchasing from the conversation itself.
Technical Guardrails: Tokens, Read/Write Separation, Verifiable IDs
On the technical side, multiple approaches are advancing in parallel.
Stripe's Andrew Beckmann, GTM lead for travel and hospitality, pointed to the Agentic Commerce Protocol (ACP) co-developed with OpenAI and Google's Universal Commerce Protocol (UCP). Stripe is also building a "Shared Payment Token" — a single-use token created at the start of a transaction and pre-scoped with limits on amount, validity window and allowed uses. The result is that sensitive credentials never need to leave Stripe even as agents act on behalf of users.
Hopper Technology Solutions (HTS) goes further. Jo Lai, HTS's SVP of AI solutions, explained: "In most use cases, an agent can search for flights, surface prices, build a cart, etc. — those are read-only actions. But when it needs to book something or process a payment, that's a write operation, and many of our partners are asking us for explicit human authorization at that step."
To enable this, HTS uses device-verified payment flows that let users complete payment outside the AI conversation, abstracting the payment action away from the LLM entirely.
Heather Dahl, CEO of Indicio, argues that identity is the missing piece. "A customer must be identifiable in a way that provides assurance that they're not a synthetic identity or a deepfake. The AI agent must be identifiable in a way that provides assurance that it's not a spoof agent." Verifiable credentials, she says, can serve as cryptographic proof tied to a verified identity.
The Behavioural Gap: Pitching Buying at the Wrong Moment
Even harder than the technology is the consumer-behaviour side. According to Mamet, consumers using AI for travel are not in a "ready to buy" state.
"Trip planning is a high-effort, long-term activity. There's often one person coordinating hotels, flights and experiences for a group. That person is already using AI to start planning. The mistake would be interrupting that planning phase with 'Book Now' calls to action before the traveler is ready," he said.
Mamet contrasts this with the Tripadvisor green-checkmark experience that finally moved the needle for Instant Booking. Asking travelers who had spent hours researching on Tripadvisor to suddenly book there was asking them to create a new habit — both a trust problem and a CVR problem.
Implications and Practical Takeaways for E-Commerce Operators
The PhocusWire analysis makes clear that for any EC operator entering agentic commerce, technical integration alone is not enough.
Bake read/write permission separation into the design. The HTS example shows that many partners now want agents to handle inventory search and price comparison while keeping payment confirmation in the human's hands. This serves both a technical and a UX checkpoint role.
Identify the moment of buying readiness. Walmart's number reflects not only low purchase rates via agents, but also potentially mistimed purchase prompts. CTAs presented to a researching consumer get ignored — designing the completion timing is the real battleground.
Create a "context of assurance" for shoppers. Just as Tripadvisor's green checkmark worked, agent-mediated purchases need visual signals and return guarantees that say "it's safe to buy here."
Conclusion
What PhocusWire's analysis surfaces is a structure where "the infrastructure is being built, but consumer habits are lagging behind." Stripe's tokens, HTS's read/write separation, Indicio's verifiable identity — these all close the technical side of the trust gap. But as Mamet repeatedly emphasises, the real battlefield is behavioural design.
What EC operators should watch next is the second-generation CVR figures from Walmart and other early movers, and whether read/write separation approaches like HTS's become the industry default. No operator will win on technology alone — operators that combine technical and consumer-context design will win the next phase.
