Expedia Reveals Travelers Distrust AI for Bookings

Expedia Group published an AI Trust Gap report showing a strong behavioral split: consumers are willing to use AI for planning but overwhelmingly prefer human-trusted brands at the point of purchase. The survey polled more than 5,700 adults across the U.S., U.K., and India and finds only 8% comfortable booking travel through AI, while 68% prefer to transact with a trusted travel brand. "Travelers don't have a technology problem with AI. They have a trust problem," said Xavi Amatriain, chief AI and data officer at Expedia Group.

The study quantifies where AI is accepted and where it is rejected. Key metrics from the survey include:

• •53% are comfortable letting AI suggest travel options
• •42% use or would use AI to monitor prices
• •40% use AI to help build itineraries
• •48% say AI saves time and helps them discover places
• •66% would not trust an AI assistant to make purchases, and only 8% would be comfortable booking via an AI platform

Expedia explicitly references integrations with ChatGPT and Gemini and highlights partnerships across OpenAI, Google, Microsoft, and Amazon. Executives including Clayton Nelson frame the move as being "wherever travelers are" while keeping the brand and its support infrastructure at the transactional moment.

This is not a technology failure so much as a product and trust problem. Consumers adopt AI as a discovery and efficiency layer but retain human-centric trust relationships for transactions that involve money, identity, and post-purchase support. The survey aligns with independent industry research showing rising AI usage for research but declining reliance on search as a default discovery tool.

For practitioners, the takeaway is pragmatic: embedding AI into UX improves discovery metrics, but moving AI into payments and bookings exposes design, security, privacy, and liability gaps. Key areas that require product and engineering focus are secure payment flows, explicit consent and approval UX, audit trails for recommendations, and fallback to human agents for exceptions.

Expect pilots that separate discovery and transaction surfaces, with AI driving personalized options but requiring explicit user authorization to execute bookings. Watch for vendor APIs and SDKs to add transaction-safe features such as consent tokens, verifiable logs, and hybrid human-in-the-loop patterns. Regulators and payment processors may push for clearer controls around data reuse and automated purchasing, which will shape implementation choices.

Implications for teams: Product managers should prioritize transparent recommendation signals and approval gates when enabling booking flows. Engineers should instrument provable data provenance and integrate robust fraud and payment protections. Data scientists should measure not only engagement gains from AI suggestions but also conversion and post-booking support costs to validate tradeoffs. Finally, legal and privacy teams must define acceptable data-sharing scopes with third-party LLMs and partners to mitigate consumer trust risks.