The query became a brief

Google's first year of AI Mode data shows travelers handing over paragraphs, not keywords — and a paragraph is an instruction to an agent

A year ago, a traveler looking for somewhere to stay in Rio typed two words into a box: "Rio hotels." Maybe three. Then they sorted by price, ticked a star rating, and scanned a list. The query was a phrase. The work happened after it.

Here is what a travel query looks like now, in Google's own example: "Help me plan a three day trip to Rio with the best places to eat like a local, must see sites, and hidden gems where I can avoid the crowds."

That is not a phrase. It is a brief — a paragraph of intent, with the priorities ranked and the conditions spelled out. The traveler is no longer telling the system what to look up. They are telling it what they want and handing over the looking.

What the data says, and what to trust in it

Google published this to mark one year of AI Mode, the AI-native version of Search it launched in the U.S. in May 2025, and the figures all point the same way. The average AI Mode query runs triple the length of a traditional one. Follow-up questions — the back-and-forth that turns a search into a conversation — rose more than 40% a month on average. Queries about planning grew 80% faster than AI Mode queries overall; brainstorming queries, 30% faster. Travel sits among the top ten topics people bring to it, and inside Google's planning tool, the most common itineraries are vacations: beach resorts, national-park hikes, honeymoons, kid-friendly trips.

A necessary caution about those numbers. They are Google's own, drawn from internal Trends data the company states plainly it measures as proportions rather than counts — there are no absolute figures behind them — and that it does not make publicly available. They arrive in a post whose purpose is to show AI Mode winning. Treat the magnitudes as Google's claims, not established fact.

What does not depend on Google's framing is the direction, because anyone can watch it happen. The query is getting longer. It is turning into a sentence a person would say out loud.

A brief is not a long keyword

The length is not the point. The change in kind is.

A keyword is a public, legible thing. "Rio hotels" means the same to every channel competing for it — the OTA, the metasearch engine, the hotel's own site, the ad system. Everyone can see it, rank against it, bid on it, measure their position in it. Two decades of distribution were built on that shared legibility: the keyword was the unit, and the whole stack was a contest to be the best answer to it.

A brief is none of those things. "Hidden gems where I can avoid the crowds" is not a term you can index against. It is qualitative, conditional, and ranked — it carries a vibe, a budget, a tolerance, a person. You cannot match it with a lookup. You have to read it, weigh the conditions against one another, and choose.

That is not retrieval. It is judgment. And judgment, performed on a paragraph of intent, on the traveler's behalf, is the definition of an agent.

The demand was already there

Which is why the demand side matters more than the supply side that took all the attention this spring.

Everything Google assembled — the cross-surface cart, the commerce protocol, the payment rail, the ads moving inside the AI answer — reads, from the outside, like a company pushing travelers toward a future they did not ask for. The query data says the opposite. The future is being pulled. Travelers are already writing briefs. A brief already needs something to read it and act on it. And the machinery Google has spent the spring building is, precisely, the thing that reads and acts.

The infrastructure is not a bet on a behavior change. It is a response to one that already happened.

The keyword era did not end because Google retired it. It ended because people stopped searching in keywords.

Losing the unit

For the hotel distribution stack, the consequence is structural — and it is not about losing a channel. It is about losing the unit.

When the query was a keyword, every layer could compete for it on common ground. When the query is a brief, the competition moves somewhere the stack cannot see. The brief gets resolved inside the agent — read, weighted, answered — before any channel is chosen. And the thing channels used to compete for, the keyword, is no longer in the query to compete for.

What is left to win is not a ranking. It is being the answer the agent assembles out of the brief. No position to hold, no term to own, no impression to count. The legible unit the whole apparatus was organized around has been replaced by one that resolves out of sight.

What changed, and what hasn't yet

So the finding in Google's report that matters is not that people are searching more, though they are. It is that they have stopped searching the way the hotel distribution system was built to be searched.

The search box did not get better at keywords. It stopped being a search box. It became a place where a traveler describes the trip they want and waits for something to arrange it — and the describing, not the arranging, is the part that already changed.

The briefs are being written. The open question is who gets to read them.

by Markus Busch, Editor/Publisher Hospitality.today

Read also: How AI Mode is changing the way people search in the U.S.

Enjoying this analysis? Hospitality.today delivers daily insights on hotel distribution, AI trends, and travel commerce — straight to your inbox. Subscribe for free at Hospitality.today →