Travel bookings in the age of ChatGPT apps: the good, the bad, and the test of the future

Diana Vorniceanu, Senior Editor at The Paypers, analyses the impact of the new ChatGPT integrations by Booking.com & Expedia on the future of travel.

AI is changing how we discover and book travel, and it's a discussion that we've been having since 2023. In 2025, all the actors in the travel industry – from travellers to airlines, hotels, to OTAs – are leveraging AI to streamline booking processes and get personalised experiences. If we talk numbers, the AI tourism market size is estimated to reach approximately USD 13.87 billion by 2030. 

However, some questions arise: does AI-assisted travel planning really solve a friction point in the booking funnel, or are these tools only creating a new level of discovery? Is the shift towards AI-enabled booking tools responding to a real customer need? This article provides an analysis of the impact the launch of ChatGPT apps might have on the travel industry and takes a look at the effect this might have on OTAs, hotels, and airlines. 

Travel apps in ChatGPT: Booking.com and Expedia, with more to follow suit

On October 6, 2025, during its much-awaited DevDay, OpenAI revealed the launch of Apps SDK with Booking.com and Expedia as travel partners. The move essentially grants the two OTAs access to ChatGPT's over 800 million weekly active users and introduces a new distribution channel for travel companies. At least on paper, this could be one of the most disruptive events in the travel industry since OTAs displaced travel agents. 

The introduction of Apps SDK marks an inflexion point in OpenAI’s model. In essence, ChatGPT now enables third-party apps to run directly within ChatGPT conversations. This new generation of apps, with which ChatGPT users can chat within ChatGPT, was made available for users outside of the European Economic Area, Switzerland, and the United Kingdom. Therefore, some geographical limitations exist, although the company stated that it plans to bring the apps to the EU. At the moment of the launch, the partnered apps were provided by Booking.com, Expedia, Spotify, Canva, Zillow, Figma, and Coursera, with more to follow by the end of the year, including Tripadvisor.

To access these in-chat apps, ChatGPT users from the regions where the functionality is available do not need a subscription plan, meaning access is completely democratised. However, just like with all AI-enabled conversations, the risk of users being provided incorrect information still exists. In its official announcement, Expedia recognises this and advises customers to double-check the details of their trips before booking. Moreover, early users have reported errors and a somewhat unreliable booking experience.

For Booking.com and Expedia’s apps in ChatGPT, at this stage, the user has to finalise the booking on the partner’s website, and not within ChatGPT. So, the integrations with Booking and Expedia maintain a clear boundary between discovery and booking. The users can search for hotels and flights in chats using natural language, can view real-time pricing and availability, as well as explore destinations on interactive maps, but they make the payment on the OTAs' sites. 

The launch of Apps SDK is one of OpenAI's most ambitious plans: turning ChatGPT into what Altman calls an AI operating system. This is not the first time travel and ChatGPT were brought up in the same sentence, as earlier in July 2025, OpenAI launched its ChatGPT agent for its Pro, Plus, and Team users. The agent was presented as being able to do tasks on behalf of the user, including planning and booking travel itineraries. 

Apps SDK also marks a step forward for OpenAI, which previously attempted to make third-party connections but failed. In March 2023, alongside partners like Expedia, FiscalNote, Instacart, KAYAK, Klarna, Milo, OpenTable, Shopify, Slack, Speak, Wolfram, and Zapier, OpenAI launched ChatGPT plugins, which were later deprecated. This new iteration solves a previous pain point: the apps are embedded within the chat itself. 

In late October 2025, OpenAI launched Atlas, an AI web browser that features an agent mode capable of taking actions on behalf of the user, which includes booking travel arrangements. The difference between booking a trip on the in-chat apps versus on Atlas resides in the fact that, on Atlas, customers can also pay for their travel arrangements with the help of the agentic tool, without having to leave the chat to complete the transaction – users simply share their payment data and make the purchase. 

The paradox of the race to ChatGPT integration: competitive advantage or poisoned apple

At first glance, the Apps SDK partnership gives the two OTAs access to 800 million daily ChatGPT users as well as a significant competitive advantage. However, the move also acts as a double-edged sword as it threatens to transform OTAs into passive order takers and cut the leverage of cross-selling on their own platforms, growing their loyalty programmes, and advertising revenue. 

In this new equation of AI-assisted travel booking, other travel industry players like airlines or hotel chains are cut out entirely from the booking process and made even more dependent on the OTAs. It is worth pointing out that, so far, no airline or hotel chain has obtained a direct ChatGPT integration, but rather remained accessible through Expedia and Booking.com, a situation that indicates a power shift. With ChatGPT funnelling all traffic through OTA partners, airlines and hotels could find themselves in the predicament of having to pay commissions on bookings they previously could capture through direct channels. 

Sanjay Vakil, Cofounder and CEO of DirectBooker, a startup that helps hotels work independently of OTAs by connecting them directly to AI platforms, offers a technical explanation for why major hotel chains weren’t among the ChatGPT launch partners, despite having comparable technology infrastructures: ‘First, in the short term, Expedia and Booking provide immediate access to about 600k hotels. By connecting to them, OpenAI is able to handle questions like "find me a hotel in Berlin" and have access to a comprehensive list of hotels rather than (only) the hotels that are part of Marriott or Hilton. What I learned at Google is that this is critically important to consumers; it is half of the reason I left to start DirectBooker.ai. 

Second, in the longer term, aggregation remains essential for scalability, efficiency, and trust. Without it, AI queries could overwhelm hotel systems with speculative searches — the “look-to-book” problem. MCP standardises data sharing but doesn’t solve infrastructure constraints; hotels’ systems aren’t built for high query volumes. Aggregators mitigate this by caching, normalising, and consolidating data feeds, enabling quick, low-load access to comprehensive and accurate inventory.’

Vakil’s assessment highlights real technical challenges, though whether the infrastructure barriers presented will become a permanent issue remains to be seen. With AI-mediated bookings still in the early stages, hotels and rentals still have a critical adjustment period before first-mover advantages become insurmountable. 

Moreover, another issue that arises has to do with how OpenAI will eventually monetise the in-chat apps and, more pressingly, how/if the costs will be trickled down to the OTAs’ travel suppliers. With commissions for hotels selling via OTAs already ranging from 15 to 30%, adding an additional commission layer could significantly thin their margins. The question is especially relevant if we take into account the rising customer acquisition costs (CAC) that the travel industry is facing. As per a recent survey, respondents from the hotel (28%), airline (26%), dining (23%), OTA (19%) and cruise (4%) sectors stated that between 2022 and 2025, these costs rose by roughly 35%, while lifetime value growth remains minimal. 

Airlines could also see similar margin compression. While the research is still in the early stages, a recent analysis spanning over 80 airlines indicates that ChatGPT traffic converts to bookings for airlines at higher rates than Google. However, the same source explains that, while the names of airlines are mentioned in ChatGPT conversations, the links shown to customers willing to book are, more often than not, those of airline competitors such as OTAs and aggregators due to insufficient citable content. 

The odd one out: Airbnb

In a recent interview where Airbnb’s CEO discussed the launch of the personalised version of the company’s AI-powered customer support function, Brian Chesky explained the launch of the chatbot as a way of doing things differently than other companies by not investing in an AI agent that aids users in the travel booking and discovery phase, but rather investing and advancing its customer support chatbots. 

In the same interview, Chesky discussed why Airbnb declined to integrate with ChatGPT by stating that the SDK ‘wasn’t quite robust enough for the things we wanted to do’. In a separate interview with Bloomberg, Chesky added that, given that Airbnb functions as a community of verified members, OpenAI would need to develop a platform that would allow Airbnb’s app to function within ChatGPT’s chatbot in a nearly self-contained way. 

The chimera of AI-assisted travel – discovery vs booking

At the moment of writing this article, studies show a dichotomy between planning for and booking a holiday using AI. According to Travala, 40% of global consumers have already used AI-based tools when planning a trip. For the UK market alone, a recent study revealed a yearly increase of 61% in the use of AI-powered platforms for travel, with younger generations (aged 18-34) being responsible for +183% of the growth. Adoption of AI use for travel is also extending to other age groups, with over the past 12 months, 48% of UK consumers aged 55 and over reportedly using AI for travel for the first time.

For the US market, Adobe reports a 3,500% YOY increase in AI traffic to US travel websites as of July 2025. The company’s survey of 5,000 US consumers also shows that nearly one-third (29%) have turned to AI tools to help plan their travels. The study reveals that consumers use AI for the research phase prior to making a decision to purchase. However, as Adobe’s research also highlights, while conversion rates from AI traffic remain lower than traditional sources, the gap is closing: in July 2025, traffic from AI sources was 47% less likely to convert, compared with 86% in October 2024.

While some statistics about higher conversions through AI chatbots exist, these refer to traditional chatbots, and no data is yet available about the conversion potential of the new AI-enabled discovery and booking tools. Historical trends, particularly those related to consumer trust in sharing payment information, indicate that adoption may take time. Additionally, the absence of clear regulatory guidance adds uncertainty to how this space will evolve.

Conclusions

ChatGPT’s October launch and the discussions that it sparked set the scene for a period of uncertainty regarding the future structure of travel distribution channels. At the moment, consumers using the in-chat apps must type the names of the OTAs manually to trigger the integrations. However, in the official press release, OpenAI mentions that ChatGPT ‘can also suggest apps when they’re relevant to the conversation’. Should automatic suggestions become standard, ChatGPT could become a new distribution gatekeeper.

While OTAs like Booking.com and Expedia can now leverage the competitive advantage that they gain by integrating in-chat apps, they also risk losing their unique position as AI could begin to take control of customer relationships, all while airlines and hotels find themselves excluded from the narrative altogether.

However, for now, the transformation remains speculative. As the customers’ use of AI for booking rather than solely for planning trips remains a critical issue that might persist, this transition period might give incumbent players time to adjust and adapt their strategies. The launch of ChatGPT apps has the potential to open up a Pandora's box in the travel industry. While the practical use of AI in travel becomes hard to ignore, at this moment, a hard pivot to using AI for bookings remains only a possibility. 

This article is part of The Paypers’ Travel Series, which includes contributions on topics spanning emerging trends in travel payments, fraud and security challenges, regulatory and tax impacts, risk management and forex, as well as sustainability in the travel industry. For a complete overview of all the contributions featured, click here. 

About the author

Diana Vorniceanu (Lupuleac) is a Senior Editor at The Paypers. She has an extensive background in content creation and is a graduate of Foreign Languages and Literature studies, currently specialising in payments and ecommerce. She strives to bring forward the latest trends for our readers, while investigating the ever-evolving landscape of cross-border payments, global ecommerce, payments for marketplaces and online platforms, and emerging technologies across the globe. You can reach Diana via email at diana@thepaypers.com or on LinkedIn.