Metasearch Bidding with AI: The Cost-Per-Acquisition Math Behind Google, Trivago, and Tripadvisor

Metasearch is the only channel where your direct rate and every OTA's rate sit side by side in front of a guest who is ready to book. Bid it well and it is your cheapest source of direct business. Bid it badly, or let parity slip, and you are paying to advertise Booking.com. This is the math that separates the two outcomes.

Why Metasearch Is the Most Misunderstood Line on the Marketing P&L

Every hotel marketing budget I have reviewed in the last three years has a metasearch line, and almost none of the owners who approved it could tell me what the channel actually costs them per stayed room night. That is not a criticism of the owners. The channel is deliberately opaque. It blends a click auction, a commission program, a free organic listing, and a live price comparison against every OTA that carries your inventory, and it reports the result as a single "return on ad spend" number that says nothing about cancellations, rate leakage, or the bookings you would have received anyway.

The prize is real. Google Hotel Ads alone commands roughly 55% of an estimated $8.2 billion in global metasearch spend, and Google Hotels is now credited with close to 60% of all metasearch reservations. For an independent property running a disciplined program, the effective acquisition cost lands at 7 to 12% of booking revenue, against the 15 to 25% the same hotel pays an OTA for the same guest. A documented case of an 85-room city hotel put the blended metasearch CPA at $19.40 per booking versus $38 for the average OTA commission. Halving your cost of sale on a channel that already drives a fifth of your online demand is a board-level number.

The trap is equally real. In Triptease's auction data, the direct rate is undercut by at least one OTA on 61% of impressions, and across all regions the direct price loses the comparison more than half the time. When your rate is not the lowest in the box, the guest clicks the OTA, the OTA collects the booking, and your metasearch spend has just funded a 20% commission. Hotels that are undercut less than 10% of the time convert 34% better than hotels undercut 30 to 40% of the time. Parity is not a distribution hygiene issue that sits next to the metasearch program. It is the metasearch program.

"Metasearch is the one place a guest sees your rate and Booking.com's rate on the same line. If yours is not the lowest, you did not buy a direct booking. You bought an OTA booking and paid twice for it."

The third shift is structural. Between 2024 and 2025 the two largest platforms moved in opposite directions on bidding. Google retired its commission-based bid strategies, closing them to new campaigns on April 30, 2024 and deprecating them entirely on February 20, 2025, pushing every hotel toward AI-driven Target ROAS and Performance Max for Travel Goals. trivago went the other way and eliminated cost-per-click entirely on September 1, 2025, making net CPA the only way to buy. A hotel that set its metasearch strategy in 2023 is now running two channels whose economics have inverted, and most have not re-underwritten either one.

The Channel Economics: Net CPA Across Every Metasearch Path

The only metric that allows a fair comparison across CPC auctions, commission programs, and OTAs is net cost per acquisition: total channel cost divided by stayed (not booked) room revenue, after cancellations. Everything else flatters one model at the expense of another. CPC campaigns look cheap when you ignore the clicks that never converted and the bookings that cancelled. Commission programs look expensive until you notice they charged nothing for either. The table below normalizes the current landscape to net CPA as a percentage of stayed revenue.

Three observations fall out of this table. First, the spread between a managed and an unmanaged Google program is as large as the spread between Google and an OTA. The channel does not have a cost. Your operating discipline has a cost. Second, trivago's move to a 10% global minimum net CPA puts a hard floor under the channel that is below every OTA and roughly at the bottom of Google's managed range, which makes it the simplest channel in the stack to underwrite. Third, the "free" row is doing more work than most hotels realize. Mirai's analysis of its client base found that free booking links delivered 34% of Google Hotels revenue against 66% from paid ads, and that share grows fastest for hotels with no paid campaign at all. If your booking engine is not feeding Google's free links with accurate live rates, you are leaving the single cheapest direct channel in existence switched off.

The OTA columns also deserve a harder look than the headline commission. Cloudbeds' guide to 2026 OTA commission rates puts Booking.com at 15% base rising toward 25% with visibility programs, and independent analyses argue the fully loaded cost, including cancellation churn, lost upsell, and the guest data you never receive, is closer to 30 to 35% per reservation. Cancellation alone is a large hidden line: OTA bookings at independent hotels cancelled at 21.8% versus 10.6% for direct in 2025. A metasearch booking is a direct booking, so it inherits the lower cancellation profile, which is why net CPA on a stayed basis is the only honest denominator.

The 2026 Bid Model Landscape by Platform

Each platform now forces a different bidding posture, and the choice of posture determines what an AI bid manager can and cannot do for you. This matrix is the state of play as of this writing.

Google's decision to sunset commission bidding is worth understanding in detail because it changes who carries the risk. Under commission per stay, Google absorbed cancellation risk and the hotel paid only on completed stays. Under Target ROAS, the hotel pays per click and Google's AI decides how much each click is worth based on the conversion value the hotel reports back. If the hotel reports gross booking value and not stayed value, the algorithm will happily pay up for last-minute, high-cancel traffic that looks lucrative at the moment of booking and evaporates before arrival. Cendyn's guidance on the bid strategy transition makes the same point: the quality of your conversion tracking is now the quality of your bidding.

trivago's move in the opposite direction is equally consequential. Because ranking on trivago is now a function of the commission you offer, your trivago Rating Index, and your price competitiveness against the OTAs on the same listing, a hotel sitting at the 10% floor will see materially less impression share than one at 18% or above. The right tier is a revenue management decision, not a marketing one: at 18% net CPA trivago is still cheaper than Expedia and you own the guest, but at 25% you have simply recreated an OTA with better data. The tier should move by season and by market, which is precisely the kind of continuous re-optimization that a human team running monthly reviews cannot deliver and an AI bid manager can.

Setting Net-CPA Targets by Segment: The Bid Strategy Matrix

The single most common error in hotel metasearch is a property-wide CPA target. A single number is wrong for every segment because the segments have different stayed values, different cancellation rates, and different displacement risk. A last-minute booker in a compression week is worth less than the rack rate suggests because you would have sold the room anyway. A 60-day-out midweek booking in shoulder season is worth more than its ADR because it fills a night you would otherwise have discounted. The bid matrix has to reflect that.

Booking windows have compressed sharply. Industry data now puts the typical window at 15 to 25 days, down from 30 to 45 pre-pandemic, with 0 to 6 day bookings representing 30 to 35% of reservations. Metasearch CPCs move with that curve: last-minute searches carry CPCs 30 to 50% above standard, while 60-day-plus searches run 20 to 30% below. Mobile is close to 40% of metasearch opportunities and converts at a lower rate unless the booking flow is stripped down. The matrix below is a starting framework we use with clients; the numbers are illustrative for a 120-room upper-upscale independent and should be recalibrated to your own stayed-revenue data.

The mechanics of implementing this matrix on Google are straightforward once the value feed is right. Target ROAS is the reciprocal of target CPA: a 10% net CPA target is a 10x ROAS, a 14% target is roughly a 7x ROAS. Google's system will then bid higher where it predicts a high-value conversion and lower where it does not. The lever the hotel controls is the definition of value. If the conversion value passed back is stayed revenue net of the historical cancellation rate for that segment, the algorithm is bidding on the right number. If it is gross booking value, the algorithm is bidding on fiction. Adchieve's overview of Hotel Ads bid management strategies covers the operational detail of layering bid adjustments for device, length of stay, and check-in window on top of the automated strategy.

One caution on Performance Max for Travel Goals. It is attractive because it extends reach across Search, YouTube, Display, Maps, and the Travel surface with a single direct-booking objective, and for a hotel with no paid presence it is the fastest path to volume. But it is also the least transparent. You will not see which surface produced which booking, and you cannot suppress a surface that is delivering high-cancel traffic. Our recommendation for properties above roughly 80 rooms is to run a dedicated Hotel Ads campaign on tROAS as the core and treat Performance Max as an incremental layer with its own budget cap and its own stayed-revenue reconciliation, rather than letting it absorb the whole metasearch budget.

Parity Is the Bid Multiplier Nobody Models

Every figure above assumes your direct rate wins the comparison. Most of the time it does not. Triptease's benchmarking across regions found the direct price undercut more than 50% of the time, in parity about a quarter of the time, and cheaper direct only 15 to 25% of the time. The undercuts are usually small. Half of Booking.com and Expedia undercuts are only 5% below the direct rate, and a quarter are within 2%. That is the OTA rebating a slice of its own commission, or a wholesaler's net rate leaking into a B2C channel, to win the click. The size of the undercut barely matters. What matters is that the guest sees a lower number next to a different logo.

This is why parity has to be wired into the bid engine rather than reported on a separate dashboard. A hotel that is undercut on a given date, device, and market is bidding into an auction it cannot win, and the correct bid in that auction is zero. Modern metasearch platforms increasingly offer real-time disparity monitoring; Metadesk's description of disparity monitoring inside metasearch and Triptease's Meta Price Match, which automatically matches the OTA rate on the direct listing, both exist because the industry finally accepted that parity and bidding are one problem. Triptease reported that hotels fixing undercuts quickly saw a 368% jump in conversion and 33% more bookings, which dwarfs anything achievable by tuning bid multipliers alone.

"Owners ask me what the right CPA target is. The honest answer is that the target is irrelevant until the parity rate is above 80%. Below that, the AI is optimizing the speed at which you lose."

The root causes of undercutting are rarely the OTA itself. In our audits the leaks come from three places: wholesale and bedbank contracts whose net rates reach B2C sites through reseller chains, promotional codes on the direct site that are not mirrored in the rate feed sent to metasearch, and room-type or occupancy mismatches where the OTA is showing a lower category than the direct listing. Each requires a different fix, and the fix belongs to revenue management, not marketing. We covered the wholesale leak in depth in our work on AI-driven wholesale channel control; the short version is that until those contracts are policed, no bid strategy will save the metasearch program. The same logic applies to the broader direct-versus-OTA economics we set out in the 2026 direct booking economics analysis.

What AI Bid Management Actually Does

The phrase "AI bidding" now covers three quite different capabilities, and hotels should be precise about which one they are buying. The first is the platform's own automation: Google's tROAS and Performance Max, trivago's ranking algorithm. These optimize for the platform's objective using the data you feed them, and they are only as good as that data. The second is a connectivity partner's bid layer, from vendors such as Koddi, Cendyn, Mirai, Triptease, D-EDGE, and the metasearch modules inside larger commerce platforms, which sits above the platforms and reallocates budget across them. The third is the hotel's own value model, which decides what a booking is worth before any bid is placed. Most of the return comes from the third, and it is the one hotels invest in least.

A properly built value model does five things. It converts every booking into stayed revenue using segment-level cancellation history. It applies a displacement factor by date so that compression nights are valued at their incremental contribution rather than their ADR. It adds a lifetime value increment for first-time direct guests who enter the CRM, since the second stay is the cheapest booking a hotel will ever acquire. It reads the live parity state and suppresses bids into lost auctions. And it feeds the resulting number back to each platform as the conversion value the platform's AI optimizes toward. This is the same forecasting discipline that underpins good revenue management, which is why the strongest metasearch programs we see are run jointly by the revenue and marketing leads rather than by an agency in isolation. Hotels building this capability often start with the forecasting layer; our AI Revenue Optimization & Forecasting service is designed to produce exactly the segment-level stayed-value and displacement estimates that a bid engine needs.

The measurement side matters as much as the bidding side. Metasearch is the last click before the booking engine more often than any other channel, so a last-click model will over-credit it and a first-click model will under-credit it. We addressed this in detail in our analysis of multi-touch attribution for hotels; the practical rule for metasearch is to hold out a geo or a date range periodically and measure incremental stayed revenue against the control. If the paid program is not producing bookings that the free links and organic search would not have produced anyway, the CPA number on the dashboard is meaningless.

A 90-Day Implementation Plan

The following sequence is what we run with properties that want to move from an agency-managed black box to an owned, AI-managed metasearch program. It deliberately puts parity and measurement ahead of bidding, because bidding on a broken foundation produces confident, well-optimized losses.

Two practical notes on execution. First, do not switch off the existing program while rebuilding it. Impression share on Google and rank on trivago both have momentum, and going dark for six weeks costs more to recover than it saves. Run the new structure in parallel on a subset of dates or markets and migrate as the gates are passed. Second, resist the temptation to let the agency or the platform define success. The only number that belongs in the board pack is incremental stayed room revenue divided by total channel cost, compared against the same figure for each OTA. If the metasearch line cannot beat the Expedia line on that basis, the money should move.

Where This Is Heading

Two forces will reshape metasearch over the next 24 months. The first is the shift of discovery toward AI trip-planning surfaces. Tripadvisor has already moved its strategic emphasis from pure metasearch toward a subscription and AI Trip Planner model, and Google's own AI-driven results increasingly answer the hotel question before the guest reaches the traditional price comparison. The bidding surface will fragment, and the hotels that have built a clean value model and a reliable rate feed will be able to plug into new surfaces quickly; those still running a CPC campaign from 2022 will not.

The second is the continued rise of direct. Skift Research projects that direct digital hotel channels will overtake OTAs by 2030, reaching $400 billion against $333 billion for OTA platforms, and the HEDNA, NYU Tisch and RateGain benchmark already shows direct matching OTA share at roughly 21% of bookings. Metasearch is the channel where that shift is decided one auction at a time. Every impression is a referendum on whether the hotel or the intermediary offers the better deal, and the hotel controls every variable that determines the outcome: the rate, the value model, the bid, and the feed. There is no other channel where that is true.

The math, in the end, is not complicated. A managed metasearch program at 8 to 12% net CPA, with parity above 80% and a stayed-value feed driving the AI, is the cheapest paid direct channel a hotel will ever operate. The same program with parity at 40% and gross booking value in the feed is an OTA subsidy with better reporting. The difference is not the platform, the agency, or the algorithm. It is whether the hotel has done the underwriting.

Frequently Asked Questions

Should a small independent hotel be on metasearch at all, given the management burden?

Yes, but start with the parts that carry no media risk. Google's free booking links cost nothing beyond a booking engine that supports the feed, and they are delivering a rising share of Google Hotels revenue, especially for hotels with no paid campaign. trivago's net CPA model charges only on completed stays, so a 40-room property can enter at the 10 to 12% tier with no risk of paying for clicks that never convert. Paid Google Hotel Ads on tROAS should come third, once the property has enough conversion history for the algorithm to learn from, which in practice means around 30 stayed bookings a month through the channel. Below that volume, Enhanced CPC with manual bid caps by booking window is the safer configuration.

Google retired commission bidding. Does that make Hotel Ads more expensive than it was?

Not necessarily, but it moves the risk to the hotel. Under commission per stay, Google absorbed cancellations and the hotel paid a fixed percentage on completed stays. Under Target ROAS the hotel pays per click, and the effective cost depends entirely on the quality of the conversion value being reported back. Hotels that pass stayed revenue net of segment cancellation rates, and that suppress bids where parity is broken, are generally seeing net CPAs in the 8 to 14% range, at or below the old commission programs. Hotels that pass gross booking value and ignore parity are seeing 15 to 20% or worse, because the algorithm is confidently buying high-cancel, low-parity traffic. Independent analyses suggest CPC bidding delivers 15 to 25% better return than commission did for hotels that actively manage it, and worse for hotels that do not.

Which trivago commission tier should we choose?

Treat it as a revenue management decision that changes by season and market rather than a one-time setting. The 10% floor is cheap but yields low impression share, because trivago ranks listings on the commission offered, the hotel's Rating Index, and price competitiveness against OTAs on the same listing. In need periods, where the incremental stayed value of a booking is high, stepping to 15 or 18% still undercuts every OTA and materially raises visibility. In compression periods, drop to the floor or pause; you will sell the room anyway. Never sit at 25% on a permanent basis. At that level you have recreated OTA economics with better guest data, which is an improvement but not the point of the channel.

How do we find out who is undercutting us and why?

Run a structured parity scan across Google, trivago, and Tripadvisor for at least 14 days, capturing every date, device, market, and room type where an OTA shows a lower rate than direct. Most metasearch platforms and several dedicated parity tools will do this automatically. Then trace each undercut to its source. Small undercuts of 2 to 5% from Booking.com or Expedia are usually the OTA rebating commission or applying a member rate. Larger undercuts, particularly from Agoda, Trip.com, or unfamiliar resellers, are almost always a wholesale or bedbank net rate that has leaked into B2C distribution through a reseller chain, and the fix is in the contract, not in the bid. Room-type mismatches, where the OTA displays a lower category than your direct listing for the same search, are a mapping error in the connectivity feed and are the fastest to correct.

What is a realistic budget and return for a 120-room independent?

A useful sizing rule is to allocate metasearch spend at roughly 8 to 12% of the direct online revenue you expect the channel to produce, and to expect the channel to reach 15 to 25% of online direct bookings within two quarters once parity is fixed and free links are live. For a 120-room upper-upscale property doing $9 million in rooms revenue with 35% of it direct online, that implies a metasearch program in the range of $50,000 to $90,000 a year producing $600,000 to $900,000 in stayed direct revenue at a net CPA near 10%. The same revenue through OTAs at a blended 18% commission would cost $110,000 to $160,000, with a higher cancellation rate and no guest data. The gap, roughly $60,000 to $80,000 a year plus the CRM value of the guests, is the return, and it scales with room count and ADR.