Clicks didn’t drift, they fell.

When Google shows an AI summary, users click any traditional result only 8% of the time (vs 15% with no AI). Only ~1% click a link inside the AI box. That’s structural loss of “surface area” for hotels and OTAs —ewer entry points, more winner-takes-all outcomes that force operators to win fewer but more decisive moments.(Pew, Jul 2025 Search Engine Land)

Zero-click creates a distribution lottery. If an assistant shows three hotels and one “Book” rail, the upstream source that powers that rail wins the entire micro-market. Everyone else gets zero— a “one shelf, one winner” dynamic.
Google’s travel gravity is still huge. Even under EU pressure, Google Hotels captures ~80% of metasearch bookings in Europe. Mirai analysis, Jun 2024 and industry lobbying summarised by PPC Land notes Google hotel usage across Europe jumped from 37% of hotels (2013) to 80% (2023). (PPC Land. mirai.com). If your “Official Site” rail isn’t visible at that moment, you’re effectively off the shelf.

Mental model: Yesterday’s funnel had lots of moss for demand to stick to. Today it’s Teflon. Your job in hotel world, whether you’re in New Zealand or not, is to create friction where it counts (structured data, direct price signals, one-tap pay) and speed where it hurts (drop-out-proof checkout).— design for certainty (policy clarity, visible member price) and speed (≤3 steps to pay).

OTA power didn’t disappear - it reassembled

App fortresses, Booking.com and Expedia have moved a huge slice of demand into logged-in, push-notified, card-on-file ecosystems. Booking’s app topped 100M users in 2024 and drove about 60% of total bookings. (Business of Apps, 2025). Expedia’s One Key launched with 168M cumulative members. (Expedia Newsroom; PR Newswire. Business of Apps). Once a guest is inside those ecosystems, discovery starts (and often ends) there.
Regulatory twist (EU). Booking’s designation as a DMA gatekeeper increases scrutiny, but doesn’t puncture app/loyalty moats. (EU DMA gatekeepers page, AP News. Digital Markets Act (DMA). Expect UI/labelling adjustments, not a collapse of loyalty-driven retention.
The arms race is paid for. The big four (Booking, Expedia, Airbnb, Trip.com) spent $16.8B on sales & marketing in 2023 (+20% YoY). (Phocuswire ; HospitalityNet. PhocusWire). Spend is already tilting from Google CPC toward loyalty/app and, next, assistant referral rails.

Takeaway: OTAs will concede some discovery to AI, then monetise execution (payments, cancellations, guarantees) and retention (loyalty, push). If you only optimise for clicks, you’ll lose where fulfilment and policy speed decide.

Make your feeds machine-readable and your checkout one-tap.

Where hotels actually win (and why it’s hard)

Parity is a stance, not a price. The fastest share shifts historically came from members-only undercuts (5–10%), not public rate wars; they’re legal across most parity regimes (including NZ), and they’re visible to assistants if you structure them in metasearch feeds (show “unlockable member price” on Google Hotel Ads)so the advantage is machine-legible, not just banner text.

The maths is quiet but brutal: OTA commission runs mid-teens to ~30%; OTA cancellations hover around ~50% vs ~18% direct (higher risk drag). Over multiple stays, direct LTV compounds via repeat + on-property wallet. (Thrivin Digital). Every booking moved from the 50% cancel pool to the 18% pool is a silent profit lift.
Direct’s “real” price is risk. A slightly better member rate + more flexible cancel + Apple/Google Pay will beat a larger nominal discount with friction. The guest’s total perceived risk is the real price - policy and pay reduce nerves more than adjectives do.

Operator test : Strip your perk + wallet pay for a week. If conversion doesn’t crater, your parity is theatre — not strategy. If it does, standardise that stack across every surface where you appear.

Assistants are becoming distribution layers

This is live, not theoretical. Booking’s AI Trip Planner (mid-2023) and Expedia’s conversational planners (2023 → now inside ChatGPT/Expedia app) compress planning into chat while handing off to native rails (OTA/airline/hotel) for payment, taxes, loyalty IDs so issuance, taxes and IDs stay correct. Quartz on Booking AI ; Expedia Newsroom ; Skift coverage. Skift
The API you feed is the shelf you get. If your CRS/IBE doesn’t expose member pricing, cancellation terms, and room-type semantics to metas + schema, the assistant can’t surface your advantage and defaults to a partner who can - room types and policies must be structured fields, not prose.

Action step: For the next 12–24 months in NZ and abroad, it won’t matter who writes the prettiest list. It will matter who renders the most machine-legible answer. (GHA on with unlockable member rate, Free Booking Links live, schema.org/Hotel + Offer + FAQPage implemented, wallet pay enabled).

What the data really predicts (three scenarios to trade against)

Base case (most likely, 2026–2030):
AI steals top-of-funnel clicks; OTA app/loyalty retains mid-funnel; hotels gain mid-single-digit direct share via member rates + consistent metasearch presence. OTA marketing mix tilts from Google CPC to loyalty/app + assistant referral fees; margins compress a few points. (Context: Google claims AI Overviews drive “more queries,” but click-outs remain depressed.) (Google Search). Operate on the assumption TOF shrinks and MOF/BOF execution picks the winner.
Upside for hotels (assistant fairness + parity momentum):
EU-style enforcement continues; assistant UIs show “Official Site” rails more consistently; hotels publishing structured member rates see double-digit direct uplift and a material drop in cancellation drag. (See EU DMA gatekeeper context above.) (EU Commission. Digital Markets Act (DMA). Only clean, structured data gets shown—messy feeds won’t be surfaced.
Downside for hotels (app monoculture):
OTA apps deepen value (BNPL, price-freeze, wallets); assistant “apps” default to OTA fulfilment; direct only grows where hotels match execution (policy, payments, latency). Evidence already visible in OTA in-app AI and relentless marketing spend. (Phocuswire ; Skift. PhocusWire). Counter by matching flexibility and checkout latency on brand.com or assistants will route elsewhere.

5 non-obvious levers that actually move the needle

Cancellation-parity asymmetry
Mirror OTA price, make direct cancel clearly more flexible — and surface it in your GHA feed + schema. You’ll win even at price parity because you reduced risk friction, not just sticker price. (Backed by OTA ≈50% vs direct ≈18% cancellations.) (Thrivin). Policy parity must be visible to machines, not hidden in copy.
Member-rate discoverability > member-rate existence
If your CRS doesn’t push “unlockable” member logic into Google Hotel Ads, assistants literally can’t see your advantage; you look the same as OTAs at the moment of truth. (GHA basics + Free Booking Links are table stakes.) (SiteMinder)
Wallets behave like a 2–3pt “rate cut” on mobile
Apple/Google Pay reduces perceived effort and payment anxiety, lifting conversion without discounting — crucial for NZ independents who can’t out-discount global brands.
Turn OTA arrivals into CRM oxygen
The only cheap direct booking is the next one. Script the front desk to convert OTA arrivals to members before checkout; trigger a 30-day single-use code. (Old-school? Yes. Still the highest-ROI move in NZ city hotels and resorts.)
Metasearch is the new lobby sign
Google still mediates the majority of travel discovery here; showing an “Official Site” rail with an unlockable member rate is the most capital-efficient top-of-funnel play left. (Mirai 80% share EU. mirai.com)

Numbers that should reframe your 2025–2026 plan

8% vs 15% CTR on AI vs non-AI SERPs; ~1% source clicks inside the AI box; higher abandonment on AI pages. (Pew; Search Engine Land. Pew Research Center)
$16.8B OTA marketing war chest in 2023 — money tilting from Google CPC toward loyalty/app and soon assistant distribution fees. (Phocuswire; HospitalityNet. PhocusWire)
168M One Key members at launch — your “do nothing” attrition pool. (Expedia Newsroom. Expedia)
Booking app: 100M+ users; ~60% bookings via app. (Business of Apps. Business of Apps)
OTA cancels ≈50% vs direct ≈18%. Your most profitable “discount” is the one that moves a booking from the 50% bucket to the 18% bucket. (Thrivin Digital. Thrivin Digital)

In New Zealand, SiteMinder’s 2025 report notes brand.com generated up to 60% higher revenue per booking vs other sources in 2024 — a directional confirmation of the direct-value thesis for local operators.) ( SiteMinder)

NZ Lens - the window of Opportunity

Metasearch under-penetration among NZ independents means turning on GHA + Free Booking Links and surfacing an unlockable member rate is still the fastest way to claw back share. (Assistants lean on these rails.)
Trans-Tasman reality: A meaningful slice of South Island leisure OTA demand originates from Aussie IPs. Mirror your member offer in AUD messaging inside GHA; push flexibility first, price second (Aussie weekenders respond to risk more than to tiny price deltas).
Airpoints-style value beats abstract points: tie direct to instantly redeemable F&B/parking credit if you don’t have an Airpoints partnership; NZ guests optimise today, not complex point charts.

What would make this wrong (and what to watch)

Assistant UX regresses: High-profile hallucinations or flaky pricing send users back to apps and metas. Watch Google’s own narrative (“AI drives more queries”) vs independent CTR data.
OTA app gravity strengthens: If One Key/Genius evolve into “Prime for travel” (credits, price-freeze, wallet refunds), direct must respond with execution, not adjectives.
Regulation cuts both ways: DMA forces fairness in Europe — but standardises assistant rails that may favour the biggest, cleanest data pipes. If your CRS can’t play, you’re not on the shelf.

Tape this above the Revenue Manager desk (and the front desk)

Price: Public parity; member-only 5–10% undercut.
Policy: Direct cancel > OTA cancel (state it in schema + GHA).
Pipes: Hotel + FAQ Page schema; member offers exposed in GHA; Apple/Google Pay on.
Proof: Track cancellation delta and repeat in 180 days as your ROI, not just last-click CPA.
People: Convert OTA arrivals to members before checkout; send a 30-day single-use code.

Bottom line

AI didn’t “steal your guests.”. It stole the extra two clicks you relied on.

OTAs are re-arming where they’re strongest: apps, payments, post-booking care. Your edge is everything an assistant and a human both recognise as lower-risk and higher-trust now: an unlockable, structured member rate; a visibly better policy; a one-tap wallet checkout; and a front-desk ritual that turns rented demand into owned demand. If you wire those in and keep them visible where assistants shop, direct stops being a slogan and starts being gravity.

AI Travel Assistants vs OTAs: Winning Direct Hotel Bookings in the Age of Artificial Intelligence