By integrating Anthropic’s Claude models into its “Penny” assistant, Priceline is shifting the competitive battleground among the major OTAs — from the race for paid traffic to a deeper contest: who owns the conversational booking interface itself. Marketing spend continues to rise in the meantime, while the Gulf is emerging as one of the world’s most active live laboratories for this technology.
Priceline — a Booking Holdings brand — has announced the next generation of its AI assistant “Penny,” now powered by Anthropic’s Claude models within Priceline’s own infrastructure, alongside Google Cloud and OpenAI. In its agentic form, the assistant moves travelers from planning to completed booking within a single conversation, handling complex requests through live maps, real-time inventory, and user preference learning.
On the surface, this reads as a product update. The correct interpretation, however, is that it signals a shift in where the battle between the major booking platforms is actually fought: from competing over who pays more to capture a customer at the search stage, to competing over who owns the conversation in which the customer makes the booking decision. That is a shift in the industry’s profit architecture — not merely in user experience.
Market Context
To grasp the scale of what is at stake, one must examine the economics of OTAs. The traditional model is built on buying traffic: players spend billions of dollars annually — on search engines and advertising — to attract customers and convert them into bookings. Estimates suggest that the four largest OTAs (Expedia, Booking Holdings, Airbnb, and Trip.com) spent approximately $17.8 billion on sales and marketing in 2024, with a substantial share directed toward Google — which plays the role of both publisher and competitor through Google Hotels.
This dependency is costly and fragile. In Q1 2026, Booking Holdings spent approximately $2.1 billion on marketing — up 16% year-on-year and equivalent to roughly 38% of revenue — while Airbnb increased its marketing spend to $751 million, a 33% rise, and Expedia grew its spending by 6%. The talk of AI efficiency has not yet reduced the marketing bill; it continues to climb.
A larger threat looms in the background. In late 2025, Google unveiled an agentic booking tool, insisting it had “no intention of becoming an OTA” and would not serve as the “merchant of record” — yet the announcement alone sent Booking Holdings and Expedia shares down between 4% and 7%, on concerns that the booking moment could migrate inside Google’s ecosystem, cutting off traffic to OTA platforms.
Analysis
This is where Priceline’s move comes into focus. When an OTA builds its own agentic assistant, it is not chasing a “new feature” — it is attempting to control the layer that could otherwise absorb its direct customer relationship. The strategic question is no longer “what is the cost per click and what is the conversion rate?” It is: who owns the interface where the booking happens, and who owns the data that flows from it?
The value of agentic AI lies in economics that run deeper than convenience. According to Priceline, early “Penny” users showed higher engagement and conversion, a reduction in support requests, and time savings of approximately ten minutes per trip. An Evercore ISI study ranked it as the best end-to-end booking experience among the AI travel tools tested. Each of these elements strikes at the core of the cost structure: higher conversion means a greater return on marketing spend; less support means lower operating costs; and a direct conversational relationship means — over time — less dependence on third-party purchased traffic.
But the picture is not a clean shift “from marketing to AI.” More precisely, AI has become a new front layered on top of the existing marketing front. OTAs are spending on both simultaneously: they continue buying traffic to defend their current market share, while investing in intelligent agents to secure their future position. This explains Booking Holdings’ own language in its 2026 official filings, where it speaks of meeting the customer “now through large language models” and of a vision for AI-powered agents coordinating trips. It is a war on two fronts — not a transition from one to another.
Industry Implications
For OTAs: The winners are those who own a data ecosystem and inventory depth that is difficult to replicate — which explains Priceline’s effort to build its agent on top of live inventory spanning more than 100 countries. At risk are smaller players who possess neither the data scale nor the capacity to fund a model race, and who risk becoming mere “inventory suppliers” behind an intelligent interface owned by someone else.
For hotels and airlines: If the booking moment migrates to a conversational layer controlled by a handful of platforms, the longstanding distribution dilemma deepens. A hotel already paying commissions that typically range from 10% to 30% to an OTA may find itself facing a new intermediary standing between it and the customer. The offsetting opportunity: AI that matches travelers with the most suitable offering could benefit hotels with a differentiated product — provided they invest in structuring their content and data properly.
For investors and governments: The market’s reaction to Google’s announcement signals that investors are now pricing “disintermediation risk” as a material factor in OTA valuations. Governments — particularly in the Gulf — face a structural question about the digital market: do they allow the intelligent booking layer to take shape in global hands, or do they support the emergence of a regional player that owns it?
Looking Ahead: Why the Gulf Matters in Particular
The paradox is that the Gulf region is not a bystander to this shift — it is among the world’s most advanced markets within it. Both Skyscanner and Almosafer have launched ChatGPT-integrated applications to facilitate AI-powered booking in the Middle East, particularly in the UAE and Saudi Arabia. This market is distinguished by high consumer confidence in AI and a supportive regulatory environment, making it a leading testbed for intelligent booking relative to the more cautious Western markets.
Yet regional players face the same pressure bearing down on Priceline. Wego’s position — approximately 35% of the Gulf metasearch market at the start of 2026 — did not shield it from referral traffic disruption following Google Flights’ expansion across the Middle East and North Africa, prompting it to concentrate on its app ecosystem and user retention through first-party data. The lesson is the same: those who do not own the direct customer interface remain dependent on a third party.
The Gulf player’s advantage does not lie in spending scale — it cannot compete with billion-dollar budgets — but in what the giants find difficult to replicate: Arabic-language content, local payment methods through which more than 60% of regional transactions flow, proximity to governments, and religious and domestic travel patterns. With local names such as Almosafer commanding 61% of total airline booking value through OTAs in the Saudi market in 2024, the region has a sufficient foundation to build a regional intelligent agent before a global platform absorbs that space.
What should be watched over the next 6 to 12 months? First, will marketing spend actually begin to decline as AI tools mature, or will it continue to rise? Second, will a major Gulf player launch a fully proprietary AI agent rather than settle for integrating third-party tools? Third, which layer will win the Arab traveler’s trust: the agency’s agent, the search engine’s agent, or a general-purpose agent like ChatGPT?
The Bottom Line
The question “Is the battle shifting from marketing to AI?” has a precise answer: no — it has expanded to encompass a new and more consequential front. Marketing still consumes a third of major OTAs’ revenues, but the real prize has become ownership of the conversational booking interface and the data behind it. Priceline’s move confirms that the giants have decided to build this layer themselves rather than cede it to Google or model providers. In the Gulf — where technology trust is high and the market is growing — the strategic choice is clear: either a regional player owns its AI agent outright, or it accepts becoming little more than inventory behind an interface owned by someone else. The window to make that decision is narrower than it appears.
