Online travel and agentic AI travel distribution: AI booking agents pose a long-term threat to OTA economics, while Airbnb is relatively insulated

Overview

The report examines how AI agents that can search and complete travel bookings may reshape online-travel distribution. Bernstein sees little immediate earnings impact, but concludes that a mature agent ecosystem could materially erode the competitive advantages and economics of hotel-focused OTAs, especially Booking and Expedia, while Airbnb has stronger defenses.

Core views

AI travel agents have moved from trip inspiration and research into booking and payment. Bernstein cites Instinct, Meta Muse and Google's AI Mode as products able to search for hotels, compare options and complete bookings on users' behalf. Instinct can access loyalty accounts, monitor prices and cancel-and-rebook reservations; its generic searches often drew results exclusively from Booking and Expedia, with suggested ordering resembling OTA results. Google AI Mode can complete hotel bookings through integrated hotel-chain and OTA partners, while Expedia announced a partnership with Meta Muse. These capabilities create a potential shift in the top of the travel funnel because users may not visit an OTA website before making a choice. Bernstein believes the present disruption is limited. Chat-based agents remain less effective than OTAs for complex hotel searches because maps, price sliders, amenity filters and interactive visual search are easier to use on established OTA interfaces. Agents also face compute and time costs when processing many inventory sources; in Bernstein's testing, Instinct produced a generic hotel search using Booking and Expedia in about two minutes, whereas a search for the best price across platforms took 10 minutes and encountered access difficulties. Large OTAs therefore remain useful one-stop inventory sources and can still influence users toward higher-take-rate inventory while capturing agent-driven bookings at little or no customer-acquisition cost. The report nevertheless sees long-term risks rising as agent interfaces improve, agents gain better access to detailed inventory, and supplier technology enables direct connections to AI platforms. First, agents could disintermediate OTAs by comparing OTA inventory with hotel-direct and other sources, turning OTAs into one of many interchangeable sources. Second, agents optimized for consumer preferences rather than OTA profitability could reduce OTAs' ability to steer users toward higher-commission listings. Bernstein estimates that Booking's Preferred and Preferred Plus programs add roughly 130 basis points to its standard commission rate and about 10% to revenue; agent-led filtering could make this inventory harder to prioritize. Third, rapid price comparison may expose a long tail of cheaper suppliers. Bernstein found that Booking and Expedia brands delivered the lowest US price only about 15% of the time in 2026, compared with 77% for smaller OTAs, potentially requiring OTAs to fund more discounts from take rates. The eventual revenue model of AI agents is central to the severity of this disruption. Bernstein considers user-funded models the largest threat because paying users are likely to demand unbiased results, while agents could list supplier inventory at zero take rates. Transaction-fee models create intermediate risk. Advertising-funded agents are the most favorable outcome for OTAs because OTAs could pay for placement and continue absorbing conversion risk, supporting their ability to monetize traffic. The report expects the eventual outcome may blend subscriptions, transaction fees and advertising, so consumer willingness to pay and tolerance for sponsored results will matter. Inventory breadth, data depth and the cost of accessing inventory determine whether agents can become compelling travel-search products. Branded hotels account for less than 10% of global room supply, so agents need independent-hotel access as well as hotel-chain partnerships. OTAs currently provide broad supply, rich hotel content, reviews and payments infrastructure. However, direct-connection providers are developing ways for independent hotels to connect to AI products at competitive commission rates. As compute and processing costs fall and agents integrate with supplier technology, Bernstein expects the protection provided by OTA scale to diminish. Bernstein does not expect Booking or Expedia to block agent access because doing so could exclude an important new booking channel while rival suppliers remain available to agents. Instead, it sees partnerships and limits on agents' post-booking activities as potential mitigants. Expedia's Muse partnership could give its inventory more complete and current representation and potentially secure economics. OTAs may also preserve merchant-of-record status and customer touchpoints for ancillary sales such as transport, experiences and insurance, though agents could eventually manage more of those post-booking interactions. Regulation is another uncertainty: rules could require users to complete booking details directly with suppliers, preserving some OTA contact, or could constrain platforms' ability to prevent agents acting for users, accelerating disintermediation. Booking and Expedia are considered the most exposed because they convert broad generic hotel-search traffic, face hotel suppliers eager to lower distribution costs, and have limited post-booking interactions to defend. Bernstein expects 2027 fundamentals to be strong, with DMA-driven changes to Google's search interface and easier Middle East comparisons more important near-term drivers than AI agents. Those results could lead markets to view agentic risk as overstated. But Bernstein's long-run conclusion remains cautious: OTAs may eventually give up market share or economics, and industry history suggests economics are more likely to be surrendered. Airbnb is relatively better insulated, although agents remain a net negative for its broader investment case. Its vacation-rental inventory is more differentiated and less price-comparable, hosts value Airbnb's tools and services such as Smart Pricing and AirCover, and the report estimates roughly 70% of its 9 million active listings are exclusive to the platform. Airbnb also receives 45% of traffic directly, reducing exposure to changes in paid generic search. Its proprietary reviews, photos, host ratings and extensive post-booking guest-host interaction make its platform valuable beyond the initial booking; Bernstein notes that 90% of users who book on Airbnb go on to message hosts. The company also plans an AI agent for 2027, which could offer its own richer agentic experience.

Analysis framework

Bernstein evaluates existing agent functionality and tests travel-search behavior, then traces how agents could change travel distribution through inventory access, consumer interface quality, supplier incentives, pricing transparency and agent monetization. It compares the exposure of Booking, Expedia and Airbnb using OTA commission economics, price-comparison evidence, traffic characteristics, inventory differentiation and post-booking customer relationships.

Methodology notes

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  Travel-distribution value-chain analysis

  The report assesses how agents could alter relationships among consumers, OTAs, hotels, independent suppliers and booking platforms, and how those changes may transfer economics away from OTAs.

• Industry AnalysisVolume-price decomposition

  Take-rate and pricing analysis

  Bernstein links OTA revenue exposure to commission rates, preferred-inventory mix, discount funding and the ability of agents to identify lower prices across booking channels.

• Valuation methodsEV/EBITDA valuation

  NTM+1 EV/EBITDA peer-benchmarked valuation

  The report values Airbnb, Booking, Expedia and TripAdvisor partly using next-twelve-month-plus-one-year EV/EBITDA multiples tied to sales growth, EBITDA margin and cash conversion.

• Valuation methodsP/E and PEG Valuation

  NTM+1 P/E valuation

  The report also uses next-twelve-month-plus-one-year P/E multiples in its disclosed valuation approaches.

• Valuation methodsDCF (Discounted Cash Flow)

  AI-disruption-risk scenario DCF

  Expedia's valuation includes a discounted-cash-flow scenario designed to reflect AI-disruption risk.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Airbnb (ABNB)

  Relatively less exposed to agentic travel disruption than hotel-focused OTAs.

  Strengths

  Differentiated vacation-rental supply, an estimated 70% exclusive inventory, host tools, 45% direct traffic and substantial post-booking guest-host interaction.

  Weaknesses

  AI agents remain a net negative to the broader investment case.

  Comparison

  Bernstein sees Airbnb as better insulated than Booking and Expedia because vacation rentals are less commoditized and less dependent on generic hotel-search traffic.

  Risks

  Overall travel-demand weakness, a step-change in competition, and regulation that restricts supply.

• Booking Holdings (BKNG)

  A hotel-focused OTA that Bernstein views as among the most exposed to long-term AI-agent disruption.

  Strengths

  Broad inventory, reviews, payments infrastructure and one-stop-shop utility currently make Booking useful to agents.

  Weaknesses

  Exposure to generic hotel-search traffic, possible loss of high-take-rate merchandising and price competition from smaller OTAs.

  Comparison

  Booking and Expedia delivered the lowest US price only about 15% of the time in Bernstein's comparison, versus 77% for smaller OTAs.

  Risks

  Loss of share to new OTAs, falling take rates as hotel supply consolidates, and more ADR- and take-rate-dilutive APAC growth.

• Expedia Group (EXPE)

  A hotel-focused OTA exposed to agent-led disintermediation but potentially supported by its Muse partnership.

  Strengths

  A formal Muse partnership may improve inventory detail and freshness and help capture agent-driven bookings.

  Weaknesses

  Generic hotel-search exposure, price competition and less ability to defend post-booking customer interaction.

  Comparison

  Like Booking, Expedia is viewed as more exposed than Airbnb to the shift from OTA-controlled search to agent-led selection.

  Risks

  Failure to realize cost savings, loss of US market share and lower take rates as hotel brands gain share.

• TripAdvisor (TRIP)

  Explicitly covered online-travel company included in Bernstein's ticker table.

  Risks

  Viator unit economics may be less profitable than expected, metasearch revenues may decline faster, and Google may target more of TripAdvisor's revenue streams.

Key data

• Instinct travel volumeApproximately $1 billion annualizedRoughly half is travel-related; Bernstein says this equates to around 0.25% of Booking traffic.
• Booking Preferred-program revenue upliftApproximately 10%Preferred and Preferred Plus add about 130 basis points to Booking's standard commission rate.
• Lowest-price frequency in the USBooking and Expedia approximately 15%; smaller OTAs 77%Bernstein's 2026 comparison of the frequency of the highest-ranked cheapest price.
• Branded hotel share of global roomsLess than 10%Agents require access to independent hotels to offer broad inventory.
• Airbnb exclusive inventoryApproximately 70% of 9 million active listingsBernstein's estimate supporting Airbnb's differentiated-supply advantage.
• Airbnb direct traffic45%The report cites this as reducing exposure to changes in paid generic search.
• Airbnb post-booking messaging90% of booking usersUsers message hosts after booking, creating post-booking platform interaction.

Impact & implications

Bernstein expects limited near-term financial disruption because agent volumes are small and large OTAs remain efficient inventory aggregators. The longer-run implication is more significant: if agents deliver richer interfaces, broad direct inventory access and unbiased price comparison, Booking and Expedia may lose control of customer acquisition and high-take-rate merchandising, resulting in lower margins or share. Airbnb's differentiated supply, host ecosystem and post-booking engagement provide relative protection but do not eliminate AI-related risk.

Risks

• Consumer trust in agents for complex bookings, rebooking and multi-vertical trips may develop more slowly than expected because of reliability and security concerns.
• Agent monetization may favor advertising or transaction fees rather than user-funded models, reducing the severity of OTA disintermediation and take-rate pressure.
• Current agent interfaces, inventory access costs and processing times may remain inferior to OTA search experiences for complex travel requests.
• Regulatory outcomes could either preserve supplier touchpoints or prevent platforms from restricting agent access, materially changing the pace of disintermediation.
• For covered companies, travel-demand trends, competition, market-share changes, take rates and company-specific execution remain target-price risks.

What to watch

• Whether consumers increasingly trust agents to complete bookings, rebook reservations and manage more complex travel tasks.
• Which agent revenue model prevails, particularly the balance among user-funded, transaction-fee and advertising models.
• The pace of agent UI improvement, direct supplier integrations and falling compute or inventory-processing costs.
• Whether OTA-agent partnerships preserve inventory access, merchant-of-record status and post-booking customer relationships.
• Regulation governing autonomous agents' ability to act for users and platforms' ability to restrict access.
• The relative near-term influence of DMA-driven Google Search changes versus AI agents on 2027 OTA fundamentals.