OpenAI recently made headlines with its move to deprioritize Instant Checkout in ChatGPT, but efforts to advance AI-powered travel booking show no signs of slowing.
The AI company’s pivot boosted OTA stock prices and led to broad speculation about the viability of agentic commerce. But the explanation may be simpler.
Ira Vouk, founder of Hospitality 2.0 Consulting, said OpenAI’s focus on building out its enterprise business “suggests that owning the transaction layer may not be their immediate priority.”
There isn’t one unified direction to achieve AI-driven booking, according to Vouk. “Each AI platform will make different decisions about whether to own, facilitate or bypass the transaction layer.”
OpenAI is continuing to invest in commerce within ChatGPT, recently introducing enhancements to product discovery, including visual browsing, side-by-side comparisons and more up-to-date results powered by its Agentic Commerce Protocol.
While OpenAI may focus on discovery—at least for now—Google is well-positioned to tap into ecosystems it has spent years developing.
According to Robert Cole, senior research analyst covering lodging and leisure travel for Phocuswright, “Google has the most impressive end-to-end AI tech stack.” In addition to its large language model (LLM), Gemini, Google’s assets include AI and quantum chips, cloud computing, identity via Google Wallet, payments via Google Pay and several open-source agentic standards.
The search giant continues to update its Universal Commerce Protocol (UCP), most recently with a new Cart option, loyalty features and Catalog functionality that allows agents to access real-time inventory and pricing.
While travel shopping brings unique challenges to agentic commerce, Google has the benefit of experience in the sector.
“With years of investment in travel (Hotel Ads, Flights, Availability, Rates and Inventory (ARI) integrations), they already have connectivity to ARI of over a million hotel properties,” said Vouk. “Their approach to hotel bookings with their new upcoming travel product appears more like a seamless pass-through rather than an attempt to own the transaction itself, which is consistent with their existing monetization model.”
Perplexity has taken yet another approach, experimenting more aggressively with owning the transaction layer through partnerships such as Selfbook.
The varied positioning among AI players suggests that there is not a single “end state,” said Vouk. “The future will likely be a mix of interface-led, ecosystem-led and transaction-led models.”
Multiple approaches to AI-driven travel booking
According to Phocuswright’s The AI Surge: Travel’s Fastest Behaviorial Shift in a Decade, U.S. travelers who use AI for planning leisure travel are most likely to do so via standalone AI platforms (e.g., ChatGPT, Claude). These platforms “are still best at the open-ended, synthesis-heavy parts of trip planning,” the report says. The share of travelers engaging with AI tools implemented within OTA and supplier websites/apps is significantly smaller.
While standalone AI platforms lead in planning and discovery, travel companies are focused on translating intent into bookings. Recent travel industry developments illustrate current approaches to bridging the gap between conversational interfaces and transactions.
Chinese online travel platform Fliggy, part of Alibaba Group, is among the most advanced examples. Following the launch of a suite of AI planning tools in 2025, the company said its AI interface can now process bookings directly through Alibaba’s Qwen app.
The capability is enabled by deep integration across Alibaba’s ecosystem. Through Qwen, Fliggy can connect travel booking with services such as payments via Alipay, mapping via Amap and other commerce functions across the Alibaba platform.
During the recent Chinese Spring Festival period, the Qwen app processed over 200 million AI-native shopping orders, and the company reports strong conversion rates for air, rail and attraction tickets.
“We’re encouraged to see that conversion performance in AI-driven booking flows is now comparable to—and in some cases improving versus—traditional booking flows,” Fliggy told PhocusWire.
In other developments, OptiStay, an AI-native hotel platform for independent properties, is focusing on infrastructure designed for machine-to machine transactions.
The company said it recently completed what it described as an autonomous hotel booking via ChatGPT, processed end-to-end through its platform at a single property.
According to OptiStay, an AI agent executed the full transaction flow, from querying live availability and selecting a room to authorizing payment and confirming the reservation directly in the hotel’s property management system.
The example remains early and limited in scope, but it highlights how enabling AI-driven booking may depend less on new interfaces and more on systems designed for agent-driven, rather than human-led, transactions.
Meanwhile, Custom Travel Solutions recently launched RouteStack.ai, which exposes hotel inventory to AI applications via Model Context Protocol (MCP) servers. The product, aimed at developers, enables end users to move smoothly from conversational search to pre-filled booking paths. The transaction itself is still completed within existing booking systems, but the approach aims to reduce friction between discovery and purchase.
A growing number of travel companies have also launched apps within ChatGPT, including Expedia, Booking.com, Skyscanner, Viator and hotel brands such as Hyatt and Accor, alongside infrastructure players like Lighthouse.
These apps largely surface real-time inventory, pricing and brand-controlled content within conversational interfaces, typically directing users to complete bookings through existing channels rather than handling transactions directly.
Looking ahead
While the AI-driven booking ecosystem continues to develop, fully agentic bookings—where AI systems complete transactions autonomously—remain a longer-term shift.
Vouk said the underlying components need to be integrated: “None of these are unsolved problems. The underlying technologies already exist, they just haven’t been unified into a cohesive, interoperable ecosystem for AI-driven transactions on a large scale.”
Cole pointed to trust as a critical missing layer, particularly around “who controls [agents], who they work for, what actions are permitted, and what they can do with the data.”
Until those pieces come together, AI will continue to reshape how travelers search and decide, even as booking remains largely rooted in existing infrastructure.