Are Travelers Ready to Pay and Check Out via Agentic AI Tools?

It can mean two different things. In the simpler version, AI helps search, compare options, and build the trip, but the traveler still clicks through and pays. In the more advanced version, the AI agent goes further and initiates or completes payment on the traveler’s behalf.

Van Alfen expects the first model to come first. As he explained, the early version is likely to be “basically a handover to a human,” where the AI does much of the look-to-book work, but the traveler still completes checkout. That keeps the transaction inside today’s familiar payment framework. A fully autonomous or delegated checkout is “completely different,” he said, because it forces the industry to review compliance, data security, privacy, fraud, and customer consent.

Because planning is easy to change. Payment is not. Travelers may be happy to use AI to explore destinations, compare hotels, or map out a route. But checkout brings in bigger questions: money, data security, refunds, disruptions, and who fixes things when something goes wrong.

Van Alfen said AI today is still mainly an “ask me anything” tool for discovery and exploration. That is very different from trusting it with a complex, high-value travel purchase. “If it’s low value, low impact, okay, go ahead. If it’s high value, high impact, I’m not sure,” he said. For a €5,000 family vacation, he added, travelers are much less likely to feel comfortable handing over payment credentials to an AI agent they do not know.

That also fits Expedia’s survey. Travelers may accept AI as a planning layer, but they still want control, accountability, and a trusted brand at the point of purchase.

Because companies do not want to lose ground if AI becomes a real booking channel later. If travelers start searching, comparing, and narrowing options inside AI tools, that could shift who controls discovery and who sits closest to the booking decision.

Van Alfen said the early push is partly about opportunity and partly about fear of missing out. Big tech, payment companies, and travel brands “go all in” because “they don’t want to be left behind,” he said. But he also warned that the market still has to prove customer adoption, monetization, and real travel use cases before AI-led checkout can become more than a race to be first.

That is where the real shift begins. If travelers still complete checkout themselves, most of today’s payment rules, fraud checks, authentication steps, and established payment rails can stay in place. The AI may help build the trip, but the payment still moves through systems that merchants, banks, card networks, and payment providers already understand.

If the agent starts paying on the traveler’s behalf, that changes. Then the industry has to rethink authentication, liability, fraud, privacy, compliance, and customer consent.

It also has to answer a basic but difficult question: who becomes the merchant of record, or the party responsible for payments, refunds, chargebacks, fraud management, and regulatory compliance? Van Alfen called delegated checkout “uncharted territory” and said that once the process moves beyond human checkout, “you open a can of worms.”

This also connects to a wider debate about agentic AI in hotel distribution. MCP could help AI agents connect to travel systems for discovery and distribution, while Google’s UCP is aimed at transactional workflows such as authorization, checkout, payments, confirmations, order management, and post-booking changes. But the payments problem is the same: once AI moves closer to checkout, the industry still has to decide who owns the transaction, and who carries the risk for consent, liability, fraud, refunds, and compliance.

The first risks are payment and servicing errors: the wrong payment method, the wrong amount, a failed refund, a chargeback dispute, or a traveler later saying they never approved what the AI selected. Asked what types of payment errors the industry should prepare for, and who should be responsible if something goes wrong, Van Alfen said the short answer is that “we don’t know yet” and “time will tell.” The reason, he explained, is that agentic checkout has not been tested at real scale.

There is also a second layer of risk: fraud. Van Alfen warned that fraudsters will use AI too. That means AI-led checkout could create a new fraud race, where merchants and payment providers must defend against increasingly automated abuse. His bottom line was simple: “We need guardrails.”

Not many. Expedia’s survey shows that only 8 percent of travelers feel comfortable booking through an AI platform, and two-thirds would not trust an AI assistant to buy or book on their behalf. Skift Research has reported an even lower level of readiness in the US market, finding that just 2 percent of US consumers would let an AI agent book on their behalf.

Van Alfen sees the same gap from the payments side. Trust is still low, mainstream awareness is still limited, and even access can be a barrier when some agentic tools sit behind paid subscriptions. Consumers already expect digital travel search to feel simple and close to free. That makes adoption harder, not easier.

Not fully. The simpler model, where AI helps but the traveler still pays, is much closer to today’s systems. But a fully delegated checkout needs more than better APIs. It needs new controls, clearer rules, and a way to prove that the agent is allowed to act for the traveler.

Van Alfen said delegated checkout would require “a lot of stuff” around the consumer and the agent. That includes a digital ID, clear consent from the traveler, and an agent that is certified and known to the merchant as a trusted agent. In practice, this means the merchant would need to know who the agent is, who it represents, and whether it is authorized to make the purchase.

The mandate is especially important. Van Alfen said there needs to be a “track and trace” showing that the customer gave approval for the agent to execute the transaction. That record matters if something goes wrong, including in a chargeback dispute, because the payment chain needs evidence of what the traveler approved and who is responsible.

So the issue is not only whether the technology can work. It is whether the payment, compliance, and liability framework around it is ready too. As AltexSoft’s MCP and UCP explainers also note, agentic travel will need infrastructure that connects discovery, checkout, payments, confirmations, and post-booking workflows, not just a smarter chatbot.

Probably. Corporate travel already has traveler profiles, travel policies, approval flows, and payment controls. That makes automation much easier in a structured environment.

Leisure travel is much less structured. A repeat company trip inside a policy framework is easier to automate than a family handing over a major holiday booking to an AI agent.

A lot has to change at once. Van Alfen pointed to better data, stronger controls, clearer liability, more trust, and better economics. He also raised one question the market still has not answered well: who is going to pay for all this?

That is why agentic checkout in travel still looks early. The tools are arriving before trust, rules, and business models are fully ready. For now, travel looks much closer to AI-assisted booking than true AI-led checkout.