User trust in autonomous booking systems

I’ve been spending a lot of time lately thinking about where AI actually adds value in travel — and just as importantly, where it doesn’t. Our new Expedia Group study, The AI Trust Gap, looks at how travelers across the U.S., U.K. and India are really using AI today. The short version: travelers don’t have a technology problem with AI. They have a trust problem. People are increasingly comfortable using AI for inspiration, discovery and itinerary planning. They’ll happily ask an AI assistant to suggest destinations, compare neighborhoods or build a draft trip. But when it comes time to actually book — to enter payment details, change a flight mid‑trip, or fix a hotel issue at 2am — they still turn to trusted brands with real‑world relationships, strong customer support and deep industry expertise. That’s not a limitation of model quality or features. It’s a reminder that in high‑stakes categories like travel, trust and accountability are the real differentiators. At Expedia Group, this is exactly how we’re thinking about AI: as a way to make travel more predictive, proactive and personalized, while keeping trusted brands and live agents at the center when it matters most. AI can help you discover and plan the trip; trusted platforms make sure the trip actually happens — and that someone is there to pick up the phone when things go wrong. If you’re interested in where AI is genuinely reshaping travel (and where human trust still does the heavy lifting), you can read the full article here:

When an AI agent shops on your behalf, how does anyone know it did what you actually asked?   That’s the question we teamed up with Google to answer with Verifiable Intent.   As AI agents begin booking travel, reordering supplies and making purchases for us, the traditional signals of intent start to disappear. A consumer may delegate a task hours or days before a transaction actually happens.   Verifiable Intent creates a tamper-resistant record that links a consumer’s identity, their original instructions and the outcome of a transaction. Privacy is built in so only the necessary information is shared.   Consumers, merchants and issuers can all verify that the agent acted within its mandate. And if something goes wrong, there are facts to review instead of guesswork.   Agentic commerce will only scale if people trust it.   That’s exactly the kind of trust infrastructure we’re working to accelerate through the AI Centre of Excellence we’re building in Singapore.   More details on Verifiable Intent here: https://lnkd.in/ga-pNKib

68% of travelers still prefer booking with a trusted travel brand over AI platforms. That's from Expedia Group's new AI Trust Gap study, surveying 5,700+ travelers across the US, UK, and India. On the surface, this reads as "AI isn't ready for travel booking." But I think the more interesting question is: what does trust actually mean in a high-stakes transaction? People are fine using AI to plan. 53% are comfortable letting AI suggest options. 40% would use it to build itineraries. But 66% won't let AI buy or book anything on their behalf. The hesitation isn't about whether the model can complete a booking. It's about who you call at 2am when your hotel has no record of your reservation. AI can generate an answer. It can't advocate for you. And here's what I think this means for the broader AI agent wave everyone is excited about. The real bottleneck to AI agents handling high-value transactions isn't intelligence. It's accountability infrastructure. Supplier relationships, customer support, operational recovery. That stuff took OTAs and travel brands decades to build. A new AI platform can't shortcut that. So the brands that win in this next phase won't be the ones with the best models. They'll be the ones that wrap AI capability around existing trust equity. That's a significant moat. #ravisbook #AIinTravel #TravelTech #DigitalMarketing

We’re entering a phase where AI capability is no longer the biggest challenge. Trust is. Everyone is racing to build smarter models. But the real question businesses are starting to ask is: Can users actually trust AI systems? For a long time, AI discussions focused only on performance. Better models. Faster outputs. Bigger benchmarks. But while working deeper with AI systems, one thing became clear: Intelligence without trust doesn’t scale. The future of AI won’t be decided by who builds the smartest model — but by who builds the most trustworthy experience. If you want AI adoption to actually work… this is the stack that matters most : The AI Trust Stack (From Transparency → User Experience) 1. Transparency → Users must understand how AI reaches decisions. → Without visibility, confidence collapses quickly. Includes: explainable decisions, audit logs, model disclosure, data sources, confidence scores. 2. Reliability → AI must behave consistently across scenarios. → Predictability builds long-term user confidence. Includes: stable models, redundancy, tested scenarios, fail-safes, predictable outputs. 3. Human-in-the-Loop → Humans remain part of critical decision workflows. → Oversight prevents automation risks. Includes: approval checkpoints, intervention control, review queues, feedback collection, escalation paths. 4. Privacy & Security → Trust grows when user data is protected by design. → Security failures instantly destroy adoption. Includes: access controls, anonymization, role permissions, audit trails, secure storage. 5. Adaptability → AI should evolve with users and real-world contexts. → Systems must learn safely without breaking trust. Includes: personalization, domain tuning, continuous improvement, learning loops, behavior updates. 6. Usability & Experience → The final layer where trust becomes invisible. → Good UX makes AI feel natural and dependable. Includes: intuitive interface, minimal friction, onboarding clarity, consistent UI, feedback prompts. Most people think AI adoption fails because of technology. In reality, it fails because trust was never designed into the system. Transparency builds confidence. Reliability builds belief. Experience builds adoption. And together, they turn AI from a tool into something users actually rely on. So the real question isn’t: “How powerful is your AI?” It’s: “How much do users trust it?” If you’re learning how AI systems are evolving beyond models into real-world products, this is exactly what I break down step by step in my weekly insights. Questions about O-1, EB-1A, or EB-5? Book a free consult - https://lnkd.in/gqJUQ-8X Join our Open Atlas community for visa-friendly job drops and free resume reviews - https://lnkd.in/gqVU84qW 🔔 Follow to stay updated on high-skilled immigration, jobs, and tech

68% stay with brands.  53% use AI to explore.  Discovery is open. Money is not. Expedia just published a report that anyone building AI in travel should read. A month ago, I asked a similar question here. ~50 people responded: founders, product leaders, travel advisors. I clustered the answers, and the pattern matched Expedia's data almost exactly. AI works for discovery.  53% use it to explore.  48% say it saves time. Duff Archie, for example, planned 5 days in Lima using ChatGPT and Gemini, and rated it higher than TikTok recommendations. But the moment payment enters the picture, everything flips. 66% wouldn't trust AI to book. 57% worry about data and payment privacy. As Diogo Galvão framed it: AI works where mistakes are cheap, and fails where they carry consequences. The industry sees the same thing. Kenny Totten, based on millions of flight shopping sessions, ranks planning as AI's strongest use case and calls automated booking "by far the worst use of AI, no one wants this." Vadim Shirinyan: "Tried almost every AI travel tool I've come across. Still haven't found one I'd use regularly." That's where the real split is. Planning is easy to improve. Booking is hard to delegate. And yet most AI products are built around booking and agents, exactly where trust is lowest. Users don't need one-click booking. They need reliability. They're not afraid of spending time. The path from first idea to completed booking has always been long, and it's getting longer. Google in 2014: travelers ran 12 searches across 22 sites over ~29 days before booking. Expedia in 2023: that journey now averages 71 days. It's likely even worse today. Which is why AI in travel today is more useful for businesses than for users: optimizing processes, reducing operational costs. No one wants to wait for user adoption. So the industry is trying to force it. Agree?

80% of travel executives said they would deploy AI booking agents at scale in 2026. 2% of travelers actually trust them to book. That gap is where OpenAI just stumbled. In March, OpenAI quietly pulled back from handling travel bookings directly inside ChatGPT. The same product that was supposed to replace OTAs. Now Expedia and Booking.com run as apps inside ChatGPT and handle the actual transaction. Here is why this matters for anyone in travel. Travelers do not want an AI to book their flight. They want an AI to help them research, compare, and shortlist. The moment the AI tries to take their money, trust collapses. Too many variables. What if the AI picks the wrong cabin. What if the refund policy is not flagged. What if the name misspells on the PNR. At HappyFares we see this every day. People spend 20 minutes comparing 4 fares before clicking pay. They want to feel in control of the final click. The winners in agentic travel are not going to be the AI that books for you. They are going to be the AI that gets you 90% of the way there and then hands the keys back to a trusted booking engine. Expedia and Booking figured that out. OpenAI figured it out the hard way. Google, Skyscanner, and Trip.com are next to learn. Where do you land on this? Is the last click something AI will ever own, or does it stay human forever?