The AI Illusion: Why Short-Term Rental Intermediaries Are Losing Their Moats

The debate over artificial intelligence and the future of travel intermediaries has been a pendulum of shifting opinions. About 18 months ago, consensus was beginning to form that traditional Online Travel Agencies (OTAs) were in serious danger. Shortly after, counterarguments emerged — particularly from industry figures like Mario Gavira — suggesting that giants like Airbnb, Vrbo, and Booking.com would actually be strengthened in this new era.

Recently, however, that defense has softened. The revised argument is that intermediaries will be harmed, but only slightly. It’s an enormous step back.

At a recent travel marketing summit, a speaker asked an audience of 250 executives how many had tracked at least one end-to-end “agentic” booking. One hand went up. The implication was clear: agentic disruption is overhyped. Bookings haven’t arrived yet.

But judging the future of the short-term rental (STR) industry based on today’s definition of an “agentic booking” is missing the fundamental change happening right beneath our feet.

The Battle for Discovery

Nobody really agrees on what an “end-to-end agentic booking” looks like. If the definition is: “I ask an AI for a villa in Mallorca and a confirmed booking magically appears in my inbox,” then no, we’re not there yet.

But we don’t need to be for massive disruption to occur.

The immediate threat to STR platforms is that discovery is shifting from their search bars to AI interfaces. Today, guests spend hours switching between Airbnb, Vrbo, and Booking.com, reading reviews and comparing maps. Tomorrow, they’ll simply ask an AI model to find a property that matches their exact criteria. You could have asked that same skeptical summit audience about mobile bookings in 2008. Eighteen years later, mobile is the default mode.

When discovery shifts, the intermediary loses its primary value proposition. Let’s examine the “unbreakable” moats that OTAs claim will protect them and why they fail in an AI-first world.

The Myth of Loyalty Programs

The first defense is loyalty: Booking.com’s Genius program, Expedia’s One Key, and Marriott’s Bonvoy. The assumption is that users will instruct their AI agents to prioritize these programs, maintaining OTA dominance.

This completely inverts human psychology. Loyalty programs in travel exist largely because comparing prices across dozens of property managers and platforms is exhausting. Humans accept slightly worse value in exchange for not having to think about it. Loyalty points are essentially a tax on human cognitive fatigue.

AI agents don’t get tired. They don’t suffer from tab fatigue. The entire purpose of an AI agent is to remove friction from price comparison. A user will simply say: “Find me the best 3-bedroom villa in Tuscany for under €4,000.” Loyalty programs will be reduced to their pure mathematical value. If a local property manager’s direct booking price beats the “Genius” discount, the AI will prioritize the direct booking.

The “Unique Inventory” Argument

The STR industry’s favorite defense is unique inventory. The argument assumes that because a specific medieval Scottish castle is listed only on Airbnb, AI agents will always be forced to route traffic through Airbnb.

This is a backward-looking assumption. Professional property managers and ambitious hosts are aggressively adopting multi-channel distribution and building their own direct booking websites. When AI agents begin to control consumer demand, property managers will ensure their direct booking APIs are accessible to those agents.

When demand shifts toward AI, supply will meet it directly. The agent creates instant demand for anyone willing to step outside the walled gardens of the major OTAs.

The Package Dodge

Finally, intermediaries argue that booking a trip is too complex for an AI, requiring sophisticated matching of flights, cars, and accommodations. Google’s recent “Universal Cart” announcement proves that false.

Packages as an OTA defense have always been just a dodge for the scarcity of human attention. Software’s attention is infinite. The moment AI agents are capable of creating dynamic packages, OTA packages will become rigid, expensive, and slow compared to what an AI can build on demand. An AI can seamlessly stitch together a cheap flight, an Uber from the airport, and a direct booking with an independent STR host in seconds.

The Ideal STR Booking Experience

If we remove our current technological constraints, what does the ideal STR booking look like? The AI will instantly scan the entire internet and return a highly curated subset of options. For a specific property, you’ll likely see three results:

• The Direct Booking Link: Coming directly from the property manager’s website. It offers the best price and direct communication with the host.
• The High-Trust OTA Link: A link to the exact same property on Airbnb or Booking.com, at a slightly higher price, for users who still want the platform’s perceived safety net.
• The Value Option: Perhaps a slightly different property that perfectly matches the user’s criteria but has stricter cancellation policies in exchange for a massive discount.

Once an AI has exhausted every corner of the internet to present these options, the user has no incentive to go directly to an OTA to search manually.

The Grace Period

There’s a catch. As far as the actual transaction goes, top-tier OTAs like Airbnb and Booking.com offer a remarkably smooth checkout experience and immense consumer trust.

Will AI eventually build a better experience? Yes. But will it happen fast enough to break deep consumer habits overnight? No. This grants the giants a grace period. They still have the capital and data to optimize their ad spend and maintain visibility.

But if you’re a mid-tier OTA, a channel manager without a direct booking strategy, or a platform that relies exclusively on search fatigue to drive conversions? Your slice of the STR transaction is about to disappear.

How is your property management company currently positioning its direct booking infrastructure to ensure it can be easily discovered by AI agents rather than relying exclusively on OTAs?