ChatGPT Ads vs. Google Ads: Where should your hotel invest?

Every time a new acquisition channel shows up, the same temptation follows: spend first, and figure out later how much volume it can deliver and at what cost. That is exactly what is happening with ChatGPT. But before moving any budget, it pays to look at how much demand the channel actually offers and what tools it really puts in a hotel's hands.

At Paraty Tech, we ran the numbers. We analyzed the Spanish hotel market across demand, cost and profitability, benchmarking against our own client portfolio data in Google Ads, Meta and metasearch. The takeaway is clear. Today, acquisition budget still makes more sense on Google, while presence in generative AI should be built starting now, with no paid media required.

So the opportunity is not in waiting. It is in starting today on the organic visibility that gets a hotel into AI answers.

In this article

• AI already shapes the choice. But how much demand does it drive?
• The recommendation comes first. The ad comes after
• An expensive click needs more than intent
• Opportunity cost counts too
• Conversational search can already be worked on Google
• Before you invest, there is work to do
• This decision gets reviewed, not closed
• Earn ChatGPT. Buy Google.

AI already shapes the choice. But how much demand does it drive?

A traveler can ask ChatGPT for something Google has traditionally struggled to handle: “hotel in Málaga with a heated pool, accessible, less than 20 minutes from the airport.” The answer goes beyond a list of links. It interprets complex intent, filters options and can present a shortlist of properties. And it offers something rare: a recommendation layer where a hotel can be cited without going through an OTA’s commercial filter.

On top of that, users who arrive from a conversational assistant do so after longer research, with the hotel already shortlisted and most of their objections resolved. That is potentially high-quality traffic. But that visibility cannot be bought.

Google handles roughly 35 million lodging-related searches per month in Spain. ChatGPT, around 1.6 million prompts. Comparing those two figures at face value can be misleading, because a conversational prompt and a Google search do not carry the same intent. What matters is demand with commercial intent. On Google, that means about 21 million searches (60% of the total), versus about 256,000 prompts on ChatGPT (16%).

That gap translates into a number that matters for any hotel deciding where to put its budget. ChatGPT accounts for just 1.2% of commercial hotel demand in Spain. Put the other way around, Google drives roughly 82 times more demand with purchase intent.

Let’s be upfront about the data. OpenAI does not publish volumes, so the ChatGPT figures are an estimate built from public sources. That is why we cross-checked them against Ahrefs’ methodology, which puts ChatGPT’s global volume at around 12% of Google’s. On that basis, the equivalent share would rise to 3.2%. More generous, yes. Enough to change the scale of the gap, no.

The picture does not change outside Spain either. Across the six markets analyzed (Spain, Portugal, France, the United Kingdom, the United States and Mexico), Google never drops below 86% of search share. And in none of them does ChatGPT’s ad product yet offer hotel-grade targeting or the volume an account needs.

The recommendation comes first. The ad comes after

Let’s walk through the trip from the start. First, the traveler decides where to go (inspiration: “where should I go on vacation in October?”). Next, they look for properties that fit their situation (consideration: “boutique hotels in Cádiz for couples”). Then they compare reviews, rates and availability (decision: “Hotel X reviews, price, availability”). And finally, they book (“Book Hotel X, 3 nights, 2 adults”).

ChatGPT carries more weight in the early stages of the decision journey, when dates, destination and budget are still open. Google owns the bottom of the funnel, right where the decision is made and the booking is paid for. That is precisely where advertising makes the most sense for a hotel. Today, ChatGPT still does not offer an ad format with destination and date targeting, or rate integration, that would let a hotel compete at that moment of decision.

At Paraty Tech, we ran a simple query to see it in practice: “hotels in Tenerife.” At the top, the answer appeared along with a module showing hotels, ratings and rates. At the bottom, at the end of the conversation, was the only ad slot on the page. And who held it? A travel insurance company.

That query reveals three major limitations:

• There is a single ad slot per answer. Inventory is scarce by design, and that scarcity can drive up the cost of winning the position.
• The auction is not hotel-specific. Insurers, airlines, banks and online travel agencies are all bidding. The hotel competes for ad space against advertisers with very different industries and business models.
• The ad shows up late. By the time it appears, the user has already seen the recommendation and moved further down the decision journey.

Real hotel visibility lives in the answer and in the results module, and neither one can be bought. It is earned through content, structured data and reputation. The channel delivers value in the discovery and consideration stages, but that does not make its ad space an efficient acquisition channel for a hotel today.

An expensive click needs more than intent

OpenAI recommends starting with a maximum bid of $3 to $5 per click (€2.75–€4.60) in a relevance-weighted second-price auction, so the actual cost lands below the bid. Our estimate puts the effective CPC between €1.50 and €3.50, with €2.20 as the benchmark. Across the three scenarios we modeled (€1.50 to €3.10), a click on ChatGPT costs 4 to 9 times more than the average click in a branded search campaign, and sits at the high end of non-branded search, which is already the least profitable channel in the comparison.

An expensive click can still make sense if it converts better. To test that, we modeled three ChatGPT investment scenarios and compared them with the performance of the main hotel acquisition channels. The model is based on a benchmark hotel with an ADR of €130, an average length of stay of 2.3 nights and an average booking value of €299. For ChatGPT, we used more generous conversion rates, precisely because conversational traffic may arrive more pre-qualified. Even in the base-case scenario, we assumed a higher conversion rate than non-branded search.

Most hotel cost structures need a ROAS of 6x to 8x to be profitable once management costs are factored in. Even with those generous conversion rates, ChatGPT comes in at 3.0x in the base case and 1.4x in the worst case. Even in the best case (6.0x), it barely reaches the low end of that threshold, well below metasearch (13.6x) and branded search (51.3x).

So the question is not whether the traffic is high quality. It is whether that quality justifies the cost of acquiring it.

Opportunity cost counts too

There is another cost that never shows up in the CPC: opportunity cost. The budget used to test a new channel has to come from somewhere. By our calculations, a test cycle of about €4,000 is needed to get a reliable read on return. At a pace of €65 a day, that is roughly two months of campaign time.

The catch? That budget is already working and generating return in other channels.

If the €4,000 comes out of non-branded search, ChatGPT would need to hit at least 3.7x for the shift to break even. If it comes out of metasearch, it would need to reach 13.6x. Today, there is no observed hotel return data on ChatGPT showing that it reaches even the low end of that range.

On top of that, there are barriers no budget increase can fix:

• No targeting by destination, stay dates, availability or rate.
• A single ad slot per answer.
• Competition with advertisers from across the entire travel industry.
• Not enough volume for automated bidding to learn at account level.
• No multi-account management on par with what Google offers.

These are product barriers, not budget barriers.

Conversational search can already be worked on Google

This is probably one of the key points of the analysis. The fact that ChatGPT is not yet the place to shift budget does not mean we should ignore the change in behavior it represents.

Travelers are moving away from searching with loose keywords and starting to describe what they want in a full sentence: “hotel in Cádiz with a heated pool that allows dogs for the December long weekend.” That conversational search can already be captured on Google. And that is where AI Max comes in.

It is not a standalone campaign type, but a setting inside existing Search campaigns. It extends reach into long, conversational queries without giving up what drives performance: conversion tracking, the search terms report, brand controls, locations of interest and management at scale. According to our analysis, its CPC ranges from €0.30 to €1.20, compared with the estimated €1.50–€3.50 for ChatGPT.

The difference is not that one understands natural language and the other does not. Both respond to the same shift in user behavior. The difference is everything that happens next: buying, measuring, optimizing and scaling.

That said, AI Max is not a set-it-and-forget-it button either. It requires reviewing the search terms report every week, monitoring the share of branded terms to avoid cannibalization, and reaching a minimum conversion volume so the system can learn. Once Google rolls out AI Mode as the default search experience, this setting will no longer be optional. Turning it on now means arriving with the learning curve already behind you.

Before you invest, there is work to do

So, what do we do about ChatGPT? Don’t pay to show up. Work to show up.

Our current recommendation is to allocate 0% of the paid media budget to generative AI and keep investing in the channels that already deliver volume, measurement and scalability. But zero budget does not mean zero attention. Quite the opposite. There are three workstreams worth starting right now:

1. Measure. Set up a dedicated channel in GA4 to identify traffic coming from AI assistants, monitor how their bots crawl the domain, and review the sessions and conversions it generates every month.
2. Audit. Build a set of destination-based queries and run it regularly to see when the hotel shows up, which competitors it appears alongside, and which searches it is still invisible in.
3. Optimize. Make sure structured data is complete and validated, keep the Google Business Profile listing up to date, create content that answers real guest questions, and actively manage online reputation.

For a hotel, the generative AI opportunity is not just about winning a click. It is about being part of the answer that helps the traveler decide.

This decision gets reviewed, not closed

The landscape may change, and it probably will. That is why this analysis is not meant to close the door on investment, but to set a series of objective conditions for reassessing it every quarter.
Today, only the measurement condition is met. The conversion tag and API are already available, while the rest still need to mature. That does not mean waiting a year to take another look. If the market keeps moving at this speed, the decision will need to be reviewed as often as the product changes.

Earn ChatGPT. Buy Google.

Generative AI matters, and a lot, at the moment the traveler dreams and compares. Today, however, it represents 1.2% of commercial demand, its clicks cost 4 to 9 times more than in a branded campaign and, across the scenarios analyzed, its return falls below a hotel’s profitability threshold. Paying to show up there means paying a premium to arrive late to a conversation that has already been decided higher up the page.

The strategy therefore has two fronts: buy conversational search where it can already be measured and scaled, and work to get AI assistants to recommend the hotel through content, data and reputation. When the channel matures, the data will tell us when it is worth investing. And we will be ready to see it.