Reporter |
Editor | 文姝琪
AI has become the new way OTA (online travel agency) platforms are fighting for users.
Recently, Fliggy launched its travel AI product “Fliggy Bangbang.” Compared to earlier AI features, Bangbang can handle itinerary generation, Q&A, and even order changes, hotel upgrades, check-in seat selection, airport transfer booking, immigration form filling, and invoice processing.
Over the past few months, Ctrip, Tongcheng, Meituan, and others have all rolled out AI travel assistants or similar offerings. Most platforms are taking a similar approach: Tongcheng Travel, for instance, fused its self-developed “Chengxin AI” with DeepSeek tech to launch an “AI + real-time booking” service that lets users do smart trip planning and flight/hotel reservations. Beyond consumer-facing AI interactions, AI tools for B-end merchants and AI-powered customer support are also being deployed across companies.
Fliggy actually started exploring AI early on. Back in April last year, it introduced the travel AI product “Ask Me Anything,” built on a Multi-Agent architecture focused on trip planning and product recommendations. According to official info, the big shift with Fliggy Bangbang is taking AI Agent capabilities from the information-interaction layer down to the service-fulfillment layer — letting AI tap into the platform’s existing product, order, and service systems.
For OTA platforms, AI isn’t just about chasing a trend — it signals a shift in user entry points and service models.
In today’s market, the “big three” landscape is pretty settled. Ctrip holds a strong share in the mid-to-high-end segment, Meituan dominates more down-market cities, and Fliggy rides on Alibaba’s ecosystem with synergies to tap. From a platform operations standpoint, how AI gets woven into the complex travel-booking journey directly affects who wins the user battle.
Previously, users would typically open an OTA app, search for hotels and flights, compare options, then complete the transaction. In the AI Agent model, users can skip straight to describing what they need, and AI handles the filtering, comparing, and booking — saving time and decision fatigue. Essentially, AI has become a differentiator that platforms offer beyond just pricing.
The timing of this AI push also tracks with tech trends. A Fliggy business lead told us that from an industry lens, top-tier model costs keep dropping, and the golden window for app explosion is close. On the demand side, travel spending keeps climbing steadily, but the mix is morphing — trips are more frequent and destinations more scattered. Take this summer: niche themes like heat-escape getaways, intangible heritage, and museum tours spiked hard. User needs have shifted from “where should I go” to highly specific, individual conditions.
That same lead explained that these kinds of demands have long been filtered out by search-and-shelf product formats, but AI now has the capability to leap from “chatting” to “booking” and “getting things done.” Fliggy’s data shows Bangbang’s current result-availability score is up more than 70% from the previous generation.
Compared to other OTAs, Fliggy’s AI exploration carries its own ecosystem weight. As a travel platform inside Alibaba, Fliggy connects e-commerce, payments, and membership ecosystems. On the tech side, Bangbang is built on Alibaba’s Qwen large model, with a self-built training environment and data pipeline for post-training, resulting in a travel-domain proprietary model deployed for service. With Qwen evolving, Fliggy as an in-ecosystem app naturally has to accelerate its play.
But for OTAs, the hard part of AI implementation isn’t just tech and demand.
Simply baking AI into information recommendations yields little for platforms and invites competition from various big-model products. As general-purpose models get stronger, users are already interacting with them for info, and down the road they might complete trip planning and reservations directly through AI assistants. This week, a rumor that “Doubao charges for hotel recommendations” sparked market buzz — even after official denial, it underscored how large-model apps are shaking up the hospitality and OTA space.
So for OTAs, only by using AI to link their own supply chains and transaction systems can they truly lift conversion efficiency and move the needle on bookings.
But compared to other goods, travel products are way more complex — real-time pricing, inventory, change/cancellation rules, and offline fulfillment all pile up, making AI integration trickier.
On commercial planning for current AI products, Fliggy is keeping expectations measured. The exec told us that Fliggy Bangbang is still in gray-release testing, with the platform focused on polishing the experience. Fliggy CTO Chen Ye also said at a media briefing that the key metric right now is user interaction — more specifically, “the more tokens used, the better.” The more users engage, the more transaction and efficiency metrics will follow naturally. As for paid plans, Chen said commercialization won’t be a focus for a long, long time.
