Paid Search for Hotels in an AI-Answer World

Two thirds of Google searches now end without a click, the paid click-through rate on AI-answered queries has fallen by more than half, and the cost of the clicks that remain keeps rising. The old hotel paid search playbook, a brand campaign, a non-brand campaign, and a monthly budget that goes up when RevPAR goes up, no longer survives contact with that math. This is what replaces it: a defensible case for brand spend, a real test for non-brand, a safe way to use AI-written ads, and a reallocation model that moves money to the channel that was always cheaper.

The query pool is shrinking and the auction is getting more expensive

Paid search worked for hotels for twenty years on a simple premise: a traveler typed something into Google, Google showed a page of links, and a hotel could buy a position on that page. The premise has changed. In the first four months of 2026, 68% of US Google searches ended without a click to any website, up from 60% in 2024. Google's own AI Overview now answers the question on the results page, and when it does, 83% of those searches end without a click. In the conversational AI Mode interface, Semrush puts the zero-click rate at 93%.

The naive read is that this is an organic traffic problem and paid is insulated. It is not. Seer Interactive tracked 3,119 informational queries across 42 organizations, 1.1 million paid impressions, from June 2024 to September 2025. On queries where an AI Overview appeared, paid click-through rate fell from 19.7% to 6.34%, a 68% decline. The organic CTR on the same queries fell 61%. The AI answer block occupies 60 to 80% of a mobile viewport, and everything below it, ads included, is a scroll away.

What makes this expensive rather than merely annoying is what happens to the auction. Fewer viewable ad slots and the same number of advertisers means more bidders per slot. Foundry CRO's 2026 travel benchmark records CPC growth of 18 to 25% year over year in high-demand markets, with seasonal swings of 100 to 400% between off-peak and peak. WordStream's cross-industry data puts the travel average CPC at $2.14, low relative to legal or finance, but that blended number hides the reality of a resort brand term in July, where the two largest bidders in the auction have a combined $15.6 billion annual marketing budget.

Booking Holdings spent $8.2 billion on marketing in 2025. Expedia Group spent $7.4 billion. Both companies describe performance marketing in their filings as "search engine keyword purchases" first and everything else second. A 120-room independent with a $180,000 annual digital budget is not competing with them. It is choosing which auctions to be present in, and that is a very different discipline.

Two of these numbers deserve a second look because they cut in the hotel's favor. First, the CTR collapse is concentrated on informational queries. Seer's study set was informational by design. Early travel-specific data suggests that AI-generated answers reduce hotel website click-through by 18 to 24% on informational queries but increase it by 8 to 12% on transactional ones, where the guest already knows what they want and the AI answer sends them to book. Second, Google is now placing ads inside the AI Overview and AI Mode, and in April 2026 folded hotel ad formats into standard Search campaigns via AI Max, which means hotel inventory can surface inside the answer rather than below it. The query pool is smaller. The share of it that is worth paying for is not necessarily smaller.

Brand defense: the spend you can actually justify

Every hotel paid search budget has a brand campaign, and every year someone in finance asks why the hotel is paying Google for people who already typed the hotel's name. It is a fair question and it has a specific answer.

Across North America, roughly 600,000 Google ads are currently bidding on hotel brand names. Most of them belong to OTAs and their affiliates. When a guest searches for your property by name and clicks an Expedia ad above your organic listing, the OTA collects a 15 to 25% commission on a guest who was already on the way to you. Brand terms are also the cheapest clicks in any hotel account, often under a dollar, and they convert at five to ten times the rate of generic terms. A hotel spending $300 a month on brand and recapturing four bookings at a $250 ADR has covered the media and saved the commission on top.

The mistake is treating brand as one thing. It is at least four things with different economics, and the budget case for each is different.

The distinction that matters most is the first pair of rows. If no OTA or affiliate is bidding on your exact brand term, your ad is mostly cannibalizing your own organic click, and geo-holdout tests across retail and hospitality routinely find that brand campaigns deliver 20 to 50% incrementality, with the low end of that range in uncontested auctions. One retailer that paused brand ads in test markets lost seven orders while platform reporting had claimed five times the impact. If an OTA is bidding, incrementality jumps, because the alternative to your ad is not your organic listing, it is a Booking.com ad above your organic listing.

This is where AI earns its place in brand defense. Auction insights change hourly. An OTA that bids on your name in shoulder season may drop out in peak when your rates rise and its margin thins, or the reverse. Tools like the AI agents Operto deployed this year monitor the brand SERP continuously and adjust bids only when a competitor appears, which means the hotel pays the contested price when the auction is contested and the floor price when it is not. A properly configured brand campaign in 2026 should have a bid range, not a bid, and the range should be set by who else is on the page.

"Nobody in finance objects to brand spend once they see the auction insights report. Paying eighty cents to keep Expedia from charging you fifty dollars is the easiest ROI in the marketing plan. Paying eighty cents when nobody else is bidding is charity to Google."

Non-brand: run the test before you believe the dashboard

Non-brand search is where hotels lose the most money in the AI-answer era, and the loss is invisible because Google's attribution reports it as a win. A guest searches "boutique hotels Charleston," sees the AI Overview, scrolls, clicks your ad, browses, leaves, and books three days later after a brand search. Google Ads credits the non-brand click. Your CRM credits the brand click. Your booking engine credits direct. Nobody credits the fact that the guest's cousin recommended you at dinner.

The only way to know whether non-brand spend is producing bookings that would not otherwise have happened is a holdout test. The design is not complicated. Pick two sets of markets with similar booking history and similar seasonality. Pause non-brand campaigns in one set for four to six weeks. Compare direct bookings, not clicks, between the two groups against their pre-period baseline. A grocery chain that ran exactly this test across 12 markets found a sales lift of 0% from its non-brand paid search. The campaign had been reporting a healthy ROAS for years.

For a single property, the geo split can be by feeder market: pause non-brand for searchers in Atlanta and Charlotte, keep it running for Nashville and Birmingham, and read the direct booking delta by guest origin. The numbers are small, which is why the test needs to run through a full booking window and why the result should be read against the pre-period rather than as a raw comparison. For a portfolio, split by property.

Two practical notes. First, do not run the test in peak season, when compression fills the hotel regardless and any channel looks incremental. Run it in shoulder, when the hotel actually needs the demand and a real lift is detectable. Second, the test is cheap. A hotel spending $6,000 a month on non-brand risks $3,000 of paused spend for a six-week answer that governs $72,000 a year.

Where non-brand does survive the test, it is usually in three places: long-tail destination-plus-occasion queries ("Charleston hotel with rooftop bar for anniversary"), where the AI Overview is weakest and intent is highest; queries from feeder markets where the hotel has low organic visibility; and dated queries with a check-in date in the search, which Google's travel-specific formats now handle inside Search campaigns. What almost never survives is the broad informational query. That click is gone, and buying an ad to sit beneath the answer that replaced it is not a strategy.

AI-generated ad variants: use them, but underwrite them

Google's answer to the shrinking query pool is to make the ad itself adaptive. AI Max for Search exited beta and reached general availability on April 15, 2026, after eleven months of testing. It does three things: it expands keyword matching using keywordless intent signals, it generates headlines and descriptions from your landing page and existing assets, and it sends the click to whichever page on your site it judges most relevant. Google reports that advertisers activating it see 14% more conversions at a similar cost per action, rising to 27% for accounts that had been running mostly exact and phrase match. Campaigns already using automatically created assets and campaign-level broad match are scheduled to auto-upgrade in September 2026, which is to say now.

The numbers are real and hotels should take them. The risk is also real and it lives in three places.

The first is the definition of "conversion." AI Max optimizes to whatever conversion action it is given. If that is a booking engine page view, it will find cheap page views. If it is a completed reservation with stayed revenue passed back, it will find reservations. Hotels that feed the algorithm booking value net of cancellation rate by segment get the 14%. Hotels that feed it gross booking value on a flexible-rate property get a campaign that confidently buys cancellable demand. This is the same lesson metasearch taught last year, and it applies with more force here because AI Max also controls the copy and the landing page.

The second is brand voice. Generated headlines are assembled from your site. If your site says "Book Now and Save 20%," the AI will write that in a dozen variations, and a luxury property that never discounts publicly will find itself advertising discounts. Google added AI Brief and text disclaimers in May 2026 so advertisers can see and constrain what is generated. Use them. Brand exclusions, URL exclusions, and a reviewed asset library are not optional for a property whose rate is a statement.

The third is match expansion. Keywordless matching means the campaign will show for queries you never bid on. For a Charleston hotel that can mean "Charleston hotels" (good), "Charleston SC weather October" (a wasted click under an AI Overview), or a competitor's name (an expensive click with a poor quality score). The search terms report must be read weekly, and negative keyword lists must be maintained with the same discipline as rate parity.

The practical position is to run AI Max in a separate campaign alongside a tightly controlled brand campaign, not to let it absorb brand. Brand defense needs a predictable message and a predictable landing page. AI Max needs room to explore. Mixing the two produces a brand campaign that occasionally sends your most valuable searchers to your events page with a generated headline about parking.

The reallocation model: move money to the channel that was always cheaper

The strongest argument for cutting non-brand search was never the CTR data. It is that hotels have a channel that produces bookings at a fraction of the cost, and most of them underfund it. Reactivating a past guest costs $10 to $30 against $50 or more to acquire a new one through paid search. Hotels retain 94.9% of guest-paid revenue on brand.com bookings versus 82.1% through OTAs. A CRM program with a real repeat-guest base can fill a third of rooms with guests who cost almost nothing to acquire. Every one of those guests is someone who does not need to be won in an auction against an $8 billion competitor.

The reallocation is not "cut search, fund email." It is a shift in the ratio, funded by the two categories that fail the tests above, with the proceeds split between retention infrastructure and the parts of search that are proven to work. Here is what the model looks like for a 120-room upscale independent with a $180,000 annual digital acquisition budget.

The total is unchanged. What changes is that $54,000 leaves a channel producing bookings at $140 to $280 each and lands in channels producing them at $10 to $45. On a $260 ADR and a 2.4-night stay, that is the difference between paying 22 to 45% of a booking's room revenue to acquire it and paying 2 to 7%. Run the arithmetic on the hotel's own numbers, not these, but the shape of the result will be the same at nearly every property that has never tested non-brand.

The retention line is where the money goes because it is the line with the most headroom. Most independent hotels have a guest database they mail four times a year with the same offer to everyone. A CRM program that segments by stay history, books pre-arrival upgrades, triggers win-back offers at the 11-month mark, and suppresses discounts to guests who booked at rack last time is a different asset. Email returns $30 to $40 per dollar spent in hospitality benchmarks precisely because the audience has already said yes once.

"The hotel that cuts non-brand search and puts the money into its own guest list is not retreating from marketing. It is refusing to buy the same guest twice from a company that sells her back at a 20% commission."

Hotels sizing this shift often find the harder question is not where the money goes but whether the data is good enough to spend it well: whether the CRM sees cancellations, whether the booking engine passes net value to Google, whether guest origin is captured at the level the holdout test needs. Properties beginning this work often benefit from a revenue systems review that connects the acquisition stack to the forecast - explore our AI Revenue Optimization & Forecasting service →

Implementation: a 90-day sequence

None of this requires a new agency or a new platform. It requires the hotel to make decisions in an order that most paid search programs skip.

Days 1 to 15: fix the conversion signal. Before touching a bid, confirm what Google Ads is optimizing toward. The conversion action should be a completed reservation, the value should be booking revenue net of segment cancellation rate, and cancellations should be imported back. If the booking engine cannot pass this, that is the first project, because every automated bidding decision downstream is only as good as this number.

Days 1 to 30: restructure brand. Split brand into contested and uncontested ad groups using auction insights. Set a bid range on contested terms and a floor bid on uncontested. Add brand + modifier groups for the property's revenue centers (spa, dining, weddings, meetings) with landing pages that match. Turn on continuous SERP monitoring so bids rise only when an OTA or affiliate appears.

Days 15 to 60: run the non-brand holdout. Design the feeder-market split, pull the baseline, pause, and read. Do not adjust non-brand budgets during the test. Resist the urge to end it early when the paused markets look fine in week two; the booking window is longer than that.

Days 30 to 90: build the retention program the money will fund. Segment the guest database by recency, frequency, rate paid, and stay purpose. Build four automated flows: pre-arrival upsell, post-stay review request with a direct-booking incentive, 11-month win-back, and a lapsed-guest reactivation at 18 months. Suppress discount offers from guests whose last booking was at BAR or above.

Days 60 to 90: launch AI Max in a fenced campaign. Separate campaign, capped daily budget, brand and competitor exclusions, URL exclusions for non-booking pages, AI Brief reviewed weekly. Read the search terms report every Monday. After 60 days, compare its cost per net booking against the metasearch program and scale or cut accordingly.

Day 90: reallocate. With the holdout read, move the money. Document the test so that next year's budget conversation starts from evidence rather than from last year's spend plus inflation.

What this means for owners

The AI-answer era did not make paid search useless for hotels. It made the untested parts of it expensive and the tested parts more valuable. Brand defense against OTA bidding is more important than it was, because the guest who searches your name is now a larger share of the clicks that still exist. Metasearch, where your rate sits next to theirs, is more important for the same reason. Non-brand search is not dead, but it must earn its budget with a holdout test rather than a ROAS report that Google wrote about itself. And the retention program that most hotels have been meaning to build is no longer a nice-to-have. It is where the non-brand money should go.

An owner reviewing next year's marketing plan should ask three questions. Has non-brand search ever been tested against a holdout, or only reported? Does the brand campaign know when an OTA is bidding and when it is not? And what share of direct bookings came from guests the hotel already knew? A plan that cannot answer all three is a plan that is still buying clicks in an auction that has quietly changed the rules.

Frequently Asked Questions

If zero-click searches are at 68%, should a hotel cut paid search entirely?

No. The zero-click figure is an average across every type of query, and it is heaviest on informational searches where the AI Overview answers the question outright. Transactional hotel queries, where a guest is searching a property name, a destination with dates, or a specific occasion, still generate clicks, and early travel data suggests AI answers are sending slightly more of those searchers through to book. The right response is to stop paying for the informational queries that no longer click, keep paying for the transactional ones that do, and put the difference into the guest list. A hotel that cuts everything loses brand defense, which is the one part of the program with unambiguous ROI.

How much should a hotel spend on brand defense?

Enough to hold over 90% impression share on contested brand terms and no more than the floor bid on uncontested ones. For most independent hotels that is a few hundred to a few thousand dollars a month, scaling with the number of OTAs and affiliates bidding on the name. The number is not a percentage of budget; it is set by the auction. A hotel with no competing bidders should spend very little. A resort with Booking, Expedia, Hotels.com, and three affiliates all bidding on its name in peak season should spend whatever it takes to stay on top, because the alternative is paying 15 to 25% commission on a guest who was already coming.

Is a geo-holdout test realistic for a single 80-room hotel?

Yes, with the right design. Split by feeder market rather than by geography around the hotel, pair markets with similar history, run through a full booking window in shoulder season, and read direct bookings by guest origin from the booking engine or PMS. The sample is small, so the result will have a wide confidence range, but that is fine: the question is not whether lift is 22% or 28%, it is whether it is closer to zero or closer to 40%. A four-to-six-week test costs a few thousand dollars in paused spend and answers a question that governs tens of thousands a year. If the property truly cannot get guest origin data, that gap is itself the first finding.

Should hotels opt into AI Max, or wait?

Opt in, but in a separate campaign with controls. Google's reported lift of 14% more conversions at similar CPA is credible for accounts that were running narrow match types, and campaigns using automatically created assets are being auto-upgraded in September 2026 regardless. The hotel's job is to set the conditions before that happens: a conversion action tied to net booking value, brand and competitor exclusions, URL exclusions for non-booking pages, a curated asset library so generated headlines do not invent discounts, and a weekly review of the search terms report. Do not let AI Max absorb the brand campaign. Brand defense needs a fixed message; AI Max needs room to explore.

Where does the money go if non-brand search is cut?

Three places, in this order. First, into CRM and retention: segmentation, pre-arrival upsell, win-back automation, and suppression of discounts to guests who paid full rate last time, because reactivating a known guest costs a fraction of acquiring a new one. Second, into metasearch and contested brand defense, which are the two paid channels where the hotel's rate is shown directly against the OTA's. Third, into a capped AI Max campaign and a small testing reserve, so that next year's decisions are made on evidence. What should not happen is the money simply disappearing from the marketing line, because the demand it was chasing still exists; it is just cheaper to reach through channels the OTAs cannot outbid.