Not yet, and probably not cleanly. As of 2026, AI agents book hotels through the large distributors. Google's AI Mode hands the transaction to a partner like Booking.com or Expedia to complete, and OpenAI walked back in-chat checkout for travel. So the near-term shift is a new layer forming on top of the OTAs, which changes their role more than it removes it. That is the honest answer to the question executives keep asking, and the rest of this piece is where the money and the leverage actually sit while it plays out.
The reason the question matters is money. A hotel hands an OTA somewhere between 15% and 30% of the booking, so anything that credibly routes demand around that toll gets attention fast. This is the channel-economics side of how AI is changing hotel distribution. Our guide to agentic booking for hotels and STRs covers how a property becomes bookable by an agent; this piece is about what happens to the OTAs, the commissions, and your channel mix once agents start booking on a traveler's behalf.
Key takeaway: In 2026 AI agents book through the OTAs, which makes the near-term story a new AI layer sitting on top of the distributors. The commission gap gives direct booking a real long-term incentive, but OTA scale, rising OTA share, and an unsolved trust gap make a fast bypass unlikely. Whatever scenario plays out, an operator's leverage is the same: accurate, reachable inventory across many channels, ready for a person or an agent to book.
Agents are starting to complete bookings, and today they complete them at the OTA. When Google introduced agentic booking in AI Mode, it said a traveler would complete the booking with the partner of your choice, and it named those partners as Booking.com, Choice Hotels, Expedia, IHG, Marriott, and Wyndham. The partner is the merchant of record. Google is the front door; the OTA or the brand still takes the money and owns the reservation. Booking.com's own ChatGPT integration works the same way, letting a traveler browse inside the assistant and then finish on the Booking.com platform.
The clearest sign of how early this is came when Skift reported that OpenAI had pulled back on in-chat checkout, which Skift attributed to travelers not completing purchases inside the assistant (OpenAI framed its own move as re-prioritizing discovery). Either way, the technology to show options in a chat window is here, and the booking still finished somewhere the traveler already trusted. Demand for the idea is real, though: Phocuswright found that 44% of travelers would book directly inside an AI platform and 40% would let an AI assistant handle flight and hotel bookings on their behalf. Willingness is running ahead of behavior, which is usually how these shifts go.
The whole disintermediation question runs on the commission gap. OTA commissions for hotels generally run between 15% and 30%, and Expedia's hotel commission specifically averages around 20%. Compare that to the short-term-rental side, where Airbnb's host service fee is 3% on its split model, or about 15.5% for hosts on the host-only model. Every one of those points is margin a property would rather keep, which is exactly why the idea of an AI agent booking a guest straight into a hotel, skipping the toll, is so appealing to operators.
That gap is the standing incentive behind every "AI vs OTAs" headline and the broader agentic booking OTAs question. If a trusted agent could route a traveler directly to a property and complete the booking, the economics would favor the operator and pressure the OTA's take rate. The gap is real and it is not small. The open question is whether anything can actually capture it at scale in the near term, and there the picture is far less flattering to the disintermediation case.
The OTAs enter the agentic era from a position of enormous strength, and the numbers say their grip is tightening. The top four OTAs took 96% of the sector's $58 billion in revenue in 2024, and for independent hotels the OTA share of bookings rose to 63.4% in 2025, up from 61.3% the year before. Reliance on the OTAs is climbing in the same window that AI booking is supposed to be threatening them.
There are good reasons for that. The OTAs hold the two things an AI agent needs most to make a confident recommendation: trust and reviews. An assistant asked to book a hotel will lean toward the places with deep, verifiable review histories and a checkout the traveler recognizes, and that is the OTAs' home turf. It is also why they are racing to become the layer agents ride rather than the layer agents skip. Booking.com is already inside ChatGPT, and Expedia is standing up its own AI toolkit and connectivity for partners. The near-term effect is that the agent becomes a new source of demand pointed at the same distributors, which can deepen an operator's dependence on them rather than relieve it.
See How Agent-Ready Distribution Works
It helps to hold three scenarios in mind rather than betting on one, because the honest answer is that nobody knows which will dominate or when. Each one has a different consequence for commissions and for how an operator should think about channel mix.
| Scenario | What it looks like | What happens to commissions | What it means for your channel mix |
|---|---|---|---|
| Intermediary (today) | Agents surface options and hand the booking to the OTA or brand, which stays the merchant of record | Unchanged; the OTA take rate holds at roughly 15% to 30% | Keep broad OTA coverage and a healthy direct channel, as you do now |
| Agent-supply layer (near-term) | OTAs become the rails agents ride; the AI assistant is a new front end on top of the existing distributors | Hold or rise, since a new intermediary sits in the path | Multi-channel reach and direct both matter more, because the agent reaches you through the channels you already feed |
| Partial disintermediation (later, uncertain) | Some trusted agents route travelers directly to properties for a share of bookings, around the OTAs | Downward pressure on OTA take rates; direct upside for operators | Agent-ready direct booking and accurate owned inventory become the highest-value assets you have |
The trap is treating scenario three as inevitable and soon. Agentic AI as a whole is running well ahead of its results: Gartner projects that over 40% of agentic-AI projects will be canceled by the end of 2027, citing cost, unclear value, and weak risk controls. The future of hotel distribution probably moves through scenario two for a good while before anyone finds out whether scenario three is real.
Here is the part I am most confident about, and it is the least dramatic. In all three scenarios, the operator who wins is the one whose rates, availability, and content are accurate and reachable across every channel a person or an agent might use. If agents book through the OTAs, you need to be well-represented on the OTAs. If a direct agent layer emerges, you need your own inventory to be agent-ready and correct. There is no version of this where messy, fragmented, single-channel distribution comes out ahead.
That is not a new idea, and that is the point. The case for spreading distribution across many channels rather than leaning on any one of them is the same case we made before agents entered the picture, laid out in our hotel distribution strategy guide and our argument for why OTAs alone are not enough. The agentic shift raises the stakes on accuracy and reach; it does not overturn the fundamentals. Discovery is moving too, of course, which is the other half of the story we cover in AI search optimization for hotels.
Make Your Inventory Agent-Ready
The sensible move is to prepare for scenario two while keeping the door open to scenario three, and to do it with work that pays off regardless. Keep rates and availability synchronized and accurate everywhere, so whatever reads them gets the truth. Keep your OTA presence healthy, because that is where the agents are pointing today. Build up your direct channel and make your own inventory agent-ready, so that if a direct agent layer does emerge, you are not starting from zero. And keep the channel mix broad, because breadth is what lets you benefit from the OTAs today and hedge against them tomorrow.
None of that requires a bet on a specific platform or a specific timeline, which is what makes it safe. The operators who will do well as agents mature are the ones already doing the unglamorous distribution work well, because the AI layer rewards the same accuracy and reach it always has. Whatever the OTAs become, owning clean, reachable, broadly distributed inventory is leverage you keep.
Getting inventory ready for that world, agent-readable and consistent across every channel, is exactly what Jetstream's MCP channel manager was built for. See how it keeps your rates and availability reachable wherever a guest or an agent goes looking.
Explore the MCP Channel Manager
Not in the near term. As of 2026, AI agents book hotels through the large OTAs, which remain the merchant of record. Google's AI Mode hands the transaction to partners like Booking.com and Expedia, and OpenAI walked back in-chat checkout for travel, which Skift attributed to travelers not completing purchases inside the assistant. Direct agent booking that bypasses the OTAs is possible later, but it is not how bookings complete today.
Probably not soon, and maybe not at all in a clean way. The top four OTAs took 96% of the sector's revenue in 2024, and independent hotels drew 63.4% of their bookings from OTAs in 2025, a share that is rising. AI is more likely to become a new layer of demand pointed at the OTAs than a replacement for them, at least until the trust and completion gaps around in-agent checkout are solved.
In the near term, very little. If agents keep booking through the OTAs, commissions of roughly 15% to 30% for hotels hold, and a new intermediary in the path can even add cost. The commission gap between OTAs and lower direct or host fees is the long-term incentive for a direct agent layer, but that pressure only materializes if agents start routing bookings around the OTAs at scale, which has not happened yet.
Do the distribution work that pays off in every scenario: keep rates and availability accurate and synchronized across channels, maintain a healthy OTA presence, strengthen the direct channel, make owned inventory agent-ready, and keep the channel mix broad. None of this depends on guessing which AI-distribution scenario wins, which is exactly why it is the safe move.
Both, depending on your setup. For operators over-reliant on a single channel, an agent layer that deepens OTA dependence is a risk. For operators with accurate, broadly distributed, agent-ready inventory and a real direct channel, the same shift is an opportunity to capture demand at better economics. What decides which one you get is how ready your distribution already is.