Online travel distribution and AI agents: AI travel agents pose a modest near-term OTA threat but a potentially material long-term challenge to Booking and Expedia

Overview

This industry report examines how AI agents that can search, compare and complete travel bookings could reshape online travel distribution. Bernstein concludes that the immediate impact on OTA fundamentals should be limited, but that increasingly capable agents could materially challenge Booking and Expedia over time, while Airbnb has several structural protections.

Core views

AI travel agents have moved from trip inspiration into booking and payment. Bernstein highlights Instinct, Meta Muse and Google AI Mode as products able to search travel inventory and complete hotel bookings, while OpenAI Dots has entered as another autonomous-agent contender. Current functionality includes searching booking platforms, using customer details and payment information, logging into loyalty accounts, monitoring prices, and potentially cancelling and rebooking reservations. In Bernstein's testing, Instinct frequently sourced results from Booking and Expedia; all five hotels returned in one general search came from Booking.com, and the top suggestion matched Booking's own result ordering. This means the largest OTAs remain useful one-stop sources of broad inventory, content and booking infrastructure today. The report identifies three long-run threats to OTA economics. First, agents may disintermediate branded OTA search by interrogating OTAs alongside hotel suppliers, smaller OTAs, GDSs and bed banks, then booking with the supplier offering the best combination of price, loyalty benefits and information. OTAs would become one inventory source among many rather than the customer-facing gateway. Second, agents optimizing for user preferences may not steer demand toward higher-commission hotel inventory. Bernstein estimates Booking's Preferred and Preferred Plus programs lift its blended commission by about 130 basis points and create roughly a 10% revenue uplift; less ability to direct consumers toward such inventory could weaken take rates. Third, agents can improve price discovery and expose a long tail of cheaper suppliers. Bernstein found Booking and Expedia brands offered the lowest price only about 15% of the time in the US in 2026, compared with 77% for smaller OTAs. If agents surface those rates, larger OTAs may need greater discount funding from their take rates to compete. The current agent experience is a meaningful near-term protection. Text-based agents remain less suited to complex hotel searches than established OTA interfaces with maps, sliders, visual filters and tailored search displays. In Bernstein's testing, Instinct produced a generic hotel search using Booking and Expedia in about two minutes, whereas a cross-platform best-price search took 10 minutes and had trouble accessing some booking options. Agents also face compute and time costs when reading multiple inventory sources. These constraints currently favor large OTAs, which can provide broad, structured inventory at low cost. But Bernstein expects this insulation to erode as interfaces become more visual and adaptive, processing costs decline, and direct supplier integrations develop. Agent revenue models determine how severe the disruption becomes. User-funded subscription or usage-based models are the greatest threat because paying users will likely expect unbiased results, encouraging agents to compare many inventory sources and potentially list supplier inventory at 0% take rates. Transaction-fee models create medium risks, while advertising is the most favorable model for OTAs because travel businesses can compete to pay for placement and OTAs can continue to absorb conversion risk in exchange for economics. Bernstein expects the eventual model to be a mix of subscriptions, transaction fees, advertising and cross-subsidies; consumer willingness to pay and preference for ad-free results will be important determinants of outcome. Near-term fundamental effects should be modest. Instinct has indicated that roughly half of its approximately $1 billion annualized volume is travel-related, which Bernstein says is only around 0.25% of Booking traffic. Muse is likely larger but remains US-focused and invite-only. Much current agent-driven travel demand still routes through OTAs at effectively no customer-acquisition cost, which could be a modest short-term return-on-assets tailwind. Bernstein expects the DMA-mandated redesign of Google's Search interface and easy Middle East comparisons to matter more for FY27 OTA fundamentals than current AI agents. It expects 2027 to be strong for Booking and Expedia, even as the market may interpret those fundamentals as evidence that agentic risk is overstated. The longer-term issue is a shift in control of the travel funnel. In Bernstein's framing, OTA advantages historically came from scale, brands and marketing intensity; agents could instead place more weight on the underlying value delivered to a user. Decisions may be made before a consumer reaches the OTA transaction layer, while agents could continually monitor prices and re-shop reservations after booking. Although OTAs retain valuable supply aggregation, content, payments and service capabilities, Bernstein believes they could ultimately need to surrender either market share or economics, with industry history suggesting economics are more likely to give way. Booking and Expedia are considered most exposed because they convert broad hotel-search traffic, hotels want lower-cost distribution, price comparison is central to hotel search, and hotel bookings provide relatively few post-booking touchpoints to monetize. The report does not expect them to block agent access because that could forfeit an important booking channel while rival suppliers remain available to agents. Instead, partnerships such as Expedia's arrangement with Muse could improve inventory quality, capture a greater share of agent traffic and potentially secure economics. Maintaining merchant-of-record status and restricting agents' post-booking capabilities may also preserve customer contact and create opportunities to sell connected trips, experiences or car rental. Airbnb is relatively better insulated, although Bernstein still regards AI agents as a net negative to its broader investment case. Its vacation-rental inventory is more differentiated and less suitable for pure price comparison; Bernstein estimates approximately 70% of Airbnb's 9 million active listings are exclusive to the platform. Hosts rely on Airbnb's tools and services, including smart pricing and AirCover, reducing incentives to bypass it. Airbnb also receives 45% of traffic directly, making it less exposed to disruption in paid generic search. Its post-booking relationship is stronger: Bernstein states that 90% of guests who make an Airbnb booking subsequently message the host. These service and communication touchpoints could allow Airbnb to preserve customer engagement even when an agent intermediates the initial booking. The company also plans an AI agent in 2027 with a visually rich interface. Regulation is another key uncertainty. Rules requiring consumers to personally enter payment or personal details at suppliers could preserve OTA contact with customers, though not their ability to steer them toward higher-take-rate inventory. Conversely, rules treating agents as acting on behalf of users could require platforms to permit agent access and accelerate disintermediation. Price-parity regulation is also relevant: DMA restrictions on parity clauses in the EU reduce protections against agent-led price competition, while parity clauses may offer greater insulation in the US, where agentic travel products are initially rolling out.

Analysis framework

Bernstein assesses live travel-agent products and tests their booking, search and inventory behavior. It then evaluates OTA exposure through the mechanisms of disintermediation, take-rate loss and price competition, using OTA price comparisons, commission-program economics, user-interface observations and alternative agent revenue models. The report compares Booking and Expedia's hotel-focused distribution model with Airbnb's differentiated vacation-rental model and considers supplier access, partnerships, post-booking control and regulation.

Methodology notes

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  Travel-distribution value-chain analysis

  The report traces how agents could move customer discovery and booking decisions away from OTAs toward suppliers or alternative inventory sources, affecting traffic, commissions and booking economics.

• Industry AnalysisSupply-demand framework

  Inventory access and supplier-distribution analysis

  Bernstein evaluates the breadth, detail and cost of travel inventory available to agents, and how direct supplier connections could reduce reliance on large OTAs.

• Competition & strategyEconomic Moat and Competitive Advantage

  Comparative competitive-position analysis

  The report contrasts OTA scale and hotel-search functionality with Airbnb's exclusive inventory, host services, direct traffic and post-booking engagement.

• Valuation methodsEV/EBITDA valuation

  NTM+1 EV/EBITDA multiples

  Bernstein values the covered companies partly using forward EV/EBITDA multiples linked to sales growth, EBITDA margin and cash conversion and benchmarked against peers.

• Valuation methodsP/E and PEG Valuation

  NTM+1 P/E multiples

  The report also uses forward P/E multiples in its company valuation approaches.

• Valuation methodsDCF (Discounted Cash Flow)

  AI-disruption risk scenario DCF

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

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

  Approximately 70% of 9 million active listings are estimated to be exclusive; host tools, 45% direct traffic, differentiated discovery and post-booking messaging strengthen customer relationships.

  Weaknesses

  AI agents remain a net negative to the broader investment case and may still intermediate initial bookings.

  Comparison

  More insulated than Booking and Expedia because vacation rentals are less commoditized and less dependent on price comparison.

  Risks

  Overall travel-demand decline, a step-up in competition from Google, Booking or Expedia, and regulation restricting supply.

• Booking Holdings (BKNG)

  A primary OTA exposure to long-term agentic disintermediation, take-rate pressure and price competition.

  Strengths

  Broad inventory, review content, payments and customer-service infrastructure make it an efficient current source for agents; current agentic traffic may carry little acquisition cost.

  Weaknesses

  Reliance on generic hotel search and ability to steer demand to higher-commission inventory are vulnerable if agents control discovery.

  Comparison

  Alongside Expedia, Booking is viewed as more exposed than Airbnb; Booking and Expedia only offered the lowest US hotel price about 15% of the time in Bernstein's analysis.

  Risks

  Loss of market share to new OTAs, declining take rate as hotel supply consolidates, and APAC growth that is more ADR- and take-rate-dilutive than expected.

• Expedia Group (EXPE)

  A primary OTA exposure to agentic travel disruption, with an opportunity to mitigate risk through the Muse partnership.

  Strengths

  Broad inventory can be valuable for agent search, and a formal Muse connection may improve inventory quality and traffic capture.

  Weaknesses

  Exposure to price comparison, generic hotel-search disintermediation and reduced ability to influence inventory mix.

  Comparison

  Viewed with Booking as more exposed than Airbnb; Expedia's agent partnership is a possible mitigation rather than a complete solution.

  Risks

  Failure to realize cost savings, loss of US market share and take-rate declines as hotel brands gain share.

• TripAdvisor (TRIP)

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

  Risks

  Viator unit economics may be less profitable than expected, Meta search revenue may decline faster than expected, and Google may target more Trip revenue streams.

Key data

• Instinct travel-related annualized volumeApproximately $1 billion, with roughly half travel-relatedBernstein says this equates to only around 0.25% of Booking traffic.
• Booking preferred-inventory revenue upliftApproximately 10%Preferred and Preferred Plus add about 130 basis points to Booking's standard commission rate.
• Lowest-price frequency in US hotel searchesBooking and Expedia approximately 15%; smaller OTAs 77%Bernstein's 2026 comparison of the highest-ranked cheapest price.
• Branded hotels' share of global room supplyLess than 10%Agents require access to independent hotels to offer broad search value.
• Instinct generic hotel search response timeApproximately 2 minutesThe search used Booking.com and Expedia; a cross-platform best-price search took 10 minutes and encountered access difficulties.
• Airbnb active listings9 millionBernstein estimates approximately 70% are exclusive to Airbnb.
• Airbnb direct traffic share45%Bernstein cites direct traffic as reducing exposure to disruption in paid generic-search acquisition.
• Airbnb bookings followed by host messaging90%Post-booking communication is a key source of customer touchpoints.

Impact & implications

Bernstein expects strong OTA fundamentals in 2027 to be driven more by Google Search changes and easier Middle East comparisons than by current AI agents. The longer-term implication is more consequential: if agents deliver comprehensive, fast and unbiased search across inventory sources, Booking and Expedia may face lower take rates, more discounting and weaker control of customer acquisition. Airbnb's differentiated supply and service ecosystem provide relative protection but do not fully eliminate agentic risk.

Risks

• Travel demand could decline overall.
• Competition from Google, Booking or Expedia could increase materially.
• Regulation could restrict travel supply.
• Booking could lose share to new OTAs, face lower take rates as hotel supply consolidates, or experience more ADR- and take-rate-dilutive APAC growth.
• Expedia may fail to realize expected cost savings, lose US market share or face lower take rates as hotel brands gain share.
• TripAdvisor faces risks from less-profitable-than-expected Viator economics, faster Meta-search revenue decline and greater Google competition.

What to watch

• Whether agent interfaces become faster, visual and capable enough for complex hotel search and filtering.
• The extent to which agents gain direct access to independent-hotel inventory and lower their inventory-processing costs.
• Which agent revenue model prevails, particularly user-funded models versus advertising or transaction-fee models.
• The scale and economics of Expedia's Muse partnership and other OTA-agent partnerships.
• Whether agents retain the ability to manage post-booking interactions and re-shop reservations.
• Regulatory decisions on agent access to websites, autonomous transactions and price-parity rules.
• The effect of DMA-driven Google Search changes on FY27 OTA fundamentals.