This battered travel stock is seen as an AI victim. But the data suggests otherwise

Booking Holdings trades at 15.3x forward earnings versus its 10-year average of 22.3x, implying roughly 46% upside if its historical valuation is restored, after underperforming the S&P 500 by about 37 percentage points over the past year. Despite AI and Middle East conflict concerns, Q2 room nights grew 5%, gross bookings rose 9%, revenue increased 8%, adjusted EBITDA grew 9% to $2.6 billion, and adjusted EPS rose 15%, consistent with its long-term targets. AI referrals remain below 1% of room nights, while growing direct/app bookings, a 4.7 million-property network, and merchant-of-record payments capabilities position Booking as a fulfillment partner rather than a disintermediated travel intermediary; Kayak's $457 million goodwill impairment remains the principal evidence of AI pressure on its metasearch business.

Analysis

BKNG’s key investment question is not whether AI changes travel discovery, but whether it can displace the transaction, service, fraud and cross-border payments stack. The current evidence supports a channel-mix shift rather than disintermediation: loss of low-margin metasearch traffic can coexist with stable or expanding core take rates if AI interfaces continue to outsource fulfillment. This creates a barbell within online travel: Kayak-like discovery assets face structurally higher acquisition costs, while merchant-of-record platforms with proprietary demand, payment capability and fragmented supply should gain relative bargaining power.

The near-term catalyst is a sequence of clean constant-currency booking and room-night prints that forces investors to separate FX noise and disruption-related cancellations from underlying demand. Over 1-3 months, sustained margin expansion despite incremental AI/product spending would challenge the market’s implicit assumption that traffic protection requires a permanently higher CAC. Over 6-18 months, the more important KPI is whether AI-originated bookings convert at comparable economics and whether direct/app mix continues rising; a higher direct mix would support both EBITDA durability and multiple normalization.

The principal falsification is not LLM referral share alone, which can rise benignly, but deterioration in direct mix, commission/take rate, marketing efficiency, or merchant-booking penetration. A sustained weakening in European lodging demand, renewed flight-capacity disruption, or a meaningful Google-owned end-to-end checkout product would weaken the thesis. MAR is less directly exposed to search disintermediation because of its owned loyalty base, but BKNG’s independent-property inventory is a differentiated source of supply that branded chains cannot readily replicate.

Consensus may be over-extrapolating the impairment of a discovery asset into a threat to the entire platform. However, a full reversion to historical valuation is not a base-case timing tool: structurally higher rates, geopolitical risk and uncertainty around AI-driven customer acquisition can justify part of the discount. Underwriting should therefore rely on earnings durability and relative valuation versus travel/platform peers, not an automatic historical-multiple target.

Key Decisions for Investors

• Initiate a 6-12 month long BKNG position on confirmation that next-quarter constant-currency gross bookings and room nights remain within the company’s long-term algorithm; target a rerating toward 18-19x forward EPS, with downside reassessment if direct mix declines or marketing expense grows materially faster than revenue.
• Express relative value as long BKNG / short EXPE in equal dollar beta-adjusted sizing over 3-6 months: BKNG has greater merchant/payment exposure and a larger direct-demand base, while EXPE has relatively more sensitivity to paid-search economics. Exit if EXPE demonstrates superior direct-booking growth or BKNG’s EBITDA margin compresses for two consecutive quarters.
• Do not short GOOG solely on this setup. Travel AI can be additive to Google engagement while checkout economics accrue to partners; monitor any announced Google-owned payment, servicing or post-booking workflow before treating it as a BKNG structural threat.
• Use quarterly monitoring triggers rather than headline AI metrics: direct/app booking mix, customer-acquisition cost per room night, merchant mix, take rate, cancellation trends and constant-currency guidance. A meaningful downgrade to constant-currency booking growth or a guide-down tied to European demand would be the signal to reduce exposure.