Online travel user experience and distribution Report Interpretation

Overview

This industry note examines two shifts in online-travel user experience: the emergence of AI agents able to execute travel bookings, and Google’s DMA-driven hotel-search redesign in the EEA. Bernstein sees the latter as a near-term OTA benefit, but the former as a longer-term challenge to intermediaries’ customer relationships and monetization.

Core views

Bernstein frames recent developments as opposing forces in the online-travel user-experience battle. New AI agents show that agentic hotel and restaurant booking is already workable, reviving concerns over OTA disintermediation and travel-search monetization. At the same time, Google’s DMA response in the EEA materially weakens its hotel-search product. The report believes this is immediately positive for OTAs, although it could level competition within the OTA market rather than reinforce the largest players’ historical dominance. Over the longer term, agentic AI remains a material threat and the latest launches are an incrementally negative data point for intermediaries. The report tested Instinct, a personal AI assistant that can navigate websites and applications and complete tasks for users. In a Manhattan search for two adults for October 15–19, below $500 per night with free cancellation, it returned six options with prices, total-stay costs and review scores, and recommended Arlo Midtown. It then navigated Booking.com’s checkout flow, disclosed an all-in price, a refundable deposit, cancellation terms and payment timing, and sought explicit user approval before booking. Instinct retained the authenticated Booking.com Genius session and could compare Booking.com and Hotels.com. However, response times of 5–10 minutes made it materially slower than traditional booking channels and other AI assistants. Its mostly text-based results and lack of clear itinerary-sharing or collaborative features leave it short of the richer travel-planning experience offered by OTAs. The key strategic risk is how agents monetize. Instinct is currently free for both consumers and booking recipients, but Bernstein expects a future user-subscription or hotel/restaurant-commission model. A user-funded model would present the greatest threat to intermediaries because an agent would be incentivized to prioritize the best booking outcome for the traveler, increasing direct-booking, take-rate-pressure and pricing-competition risk. More broadly, Meta’s Muse, xAI’s Grokbot and Anthropic’s Claude Agent illustrate the shift from AI as a search and recommendation tool toward a transaction and workflow layer. That could reduce the direct customer touchpoints received by OTAs and travel providers while increasing the importance of agent-optimized distribution, pricing and content. Google’s further EEA changes followed DMA requirements and a subsequent €460m fine for non-compliance; testing began in July and the redesign was fully rolled out in September. Sponsored links remain at the top, but the principal organic result is now an aggregator unit where users can select providers such as Booking.com, Expedia or Google. The unit displays only three hotels, offers no price, rating, location or other filters, and sends almost every interaction into an OTA app or website. A separate supplier unit offers direct hotel links but has even less information, no filtering, no review scores or price comparisons, and limited clarity on why hotels were selected. Traditional map-based browsing and much comparison functionality have largely disappeared; the same framework is expected for AI Overviews and AI Mode. Bernstein agrees with Google’s assessment that the redesign is a major reduction in service quality. The report argues that consumers now find it substantially harder to compare hotels by price, reviews, location, brand, amenities and availability, or refine their searches iteratively. Consumers may instead move to AI-native discovery platforms, metasearch providers such as Trivago, or direct OTA usage. Earlier DMA changes were estimated to have reduced direct-hotel traffic by roughly 30%; Bernstein believes the more restrictive redesign could shift further traffic and bookings toward intermediaries. Only eight Amsterdam properties were surfaced directly in one supplier-unit observation, while hotels could not show live pricing or review scores. OTAs are the clearest beneficiaries within Google because they gain more high-visibility search real estate, participate in the aggregator unit for free rather than through auctions, face less price-comparison functionality and may require lower discounting and customer-acquisition spending. Users are also more often pushed directly into OTA ecosystems. Yet the report finds the distribution of benefits uncertain. Traditional Google SEO and paid search rewarded Booking’s scale, conversion rates and pricing through stronger bidding power. The new rankings are said to reflect relevance and utility rather than auction economics. Across searches in ten cities, Booking.com ranked first four times, Expedia once, Hotels.com once and Google itself three times. If scale and bidding matter less, qualifying smaller OTAs could benefit disproportionately, potentially disrupting the OTA hierarchy. Airbnb is viewed as a marginal winner from removal of Google Vacation Rentals, which modestly reduces the visibility of smaller platforms and direct-booking options.

Analysis framework

Bernstein combines hands-on product testing of an AI travel agent with an examination of Google’s revised EEA hotel-search workflow. It then traces how changes in discovery, ranking, click-through paths and pricing transparency could affect OTA traffic, customer-acquisition costs, competitive positioning and longer-term distribution economics.

Methodology notes

• Competition & strategyValue chain analysis

  Travel-distribution and customer-touchpoint analysis

  The report assesses how AI agents and Google’s redesigned search flow can reallocate discovery, booking traffic and monetization among consumers, OTAs, hotels and metasearch providers.

• Valuation methodsEV/EBITDA valuation

  NTM+1 EV/EBITDA peer-multiple valuation

  The disclosure appendix values covered companies using next-twelve-month-plus-one EV/EBITDA multiples benchmarked against peers, alongside P/E multiples.

• Valuation methodsDCF (Discounted Cash Flow)

  AI-disruption risk scenario DCF for Expedia

  Expedia’s stated 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).

• Booking Holdings (BKNG)

  A major OTA that benefits from greater OTA prominence in Google’s redesigned EEA hotel search, though its scale advantage may be less decisive under relevance-based rankings.

  Strengths

  Historically superior conversion rates and pricing supported stronger bidding power and traffic share.

  Weaknesses

  The new ranking regime may diminish scale and bidding advantages.

  Comparison

  Appeared first four times in the ten-city aggregator-ranking review; the report’s table rates it Market-Perform.

  Risks

  Could lose share to new OTAs; take rate could fall as hotel supply consolidates; APAC growth could dilute ADR and take rate.

• Airbnb (ABNB)

  A marginal beneficiary from the removal of Google Vacation Rentals in Europe.

  Strengths

  Removal modestly reduces visibility for smaller platforms and direct-booking options.

  Comparison

  The report rates Airbnb Outperform.

  Risks

  A decline in travel demand, stronger competition from Google, Booking or Expedia, or regulation restricting supply.

• Expedia Group (EXPE)

  An OTA positioned to benefit from free aggregator participation and more traffic directed into OTA ecosystems.

  Strengths

  Appeared first once in the report’s ten-city ranking review.

  Weaknesses

  AI-disruption risk is sufficiently material to be incorporated in a DCF valuation scenario.

  Comparison

  The report rates Expedia Market-Perform and gives a $310 target price in its valuation appendix.

  Risks

  Failure to realize cost savings, US share loss, lower take rates as brands gain share, weaker Viator economics, or faster metasearch revenue decline.

• TripAdvisor (TRIP)

  A metasearch-related beneficiary could emerge if consumers seek comparison, pricing transparency and filtering absent from Google’s revised experience.

  Comparison

  The report rates TripAdvisor Outperform.

  Risks

  Meta-search revenues could decline faster than expected and Google could target more of TripAdvisor’s revenue streams.

Key data

• Google DMA fine€460mCited as preceding Google’s further EEA hotel-search changes.
• Instinct response time5–10 minutesMaterially slower than traditional booking channels and other AI assistants.
• Instinct hotel-search result6 optionsFor a Manhattan stay for two adults, October 15–19, below $500 per night with free cancellation.
• Arlo Midtown indicative price$425/night; $1,846 for four nights before taxes and feesInstinct’s recommended option in the report’s test.
• Booking checkout price$2,138.02 all-inShown after Instinct progressed through Booking.com checkout for Arlo Midtown.
• Direct-hotel traffic impactroughly 30% reductionEstimate for earlier DMA-related changes, before the latest redesign.
• Aggregator ranking sampleBooking.com 4; Expedia 1; Hotels.com 1; Google 3Top positions across searches in ten cities.
• Booking supplier-unit top position40%Most common top supplier in Bernstein’s sample of 20 supplier-unit searches.

Impact & implications

The report expects the DMA redesign to improve OTAs’ near-term competitive position relative to suppliers and metasearch by increasing visibility, directing traffic into OTA ecosystems and reducing auction-based acquisition costs. It cautions that AI agents could ultimately weaken intermediaries’ direct customer relationship, especially if their economics prioritize consumer outcomes over supplier commissions.

Risks

• Agentic AI could disintermediate OTAs by steering travelers to direct booking and reducing intermediary take rates.
• A user-funded AI-agent model would most strongly prioritize traveler outcomes, increasing pricing competition for intermediaries.
• The DMA redesign may disrupt OTA competitive hierarchy if relevance-based ranking reduces the value of scale and bidding power.
• Travel-demand weakness, share losses, take-rate pressure and regulation are stated downside risks for covered companies.