The BureauPlacePressCultureAll stories
Press
November 11, 2026

AI booked four Uber rides in Halloween week

AI agents booked four Uber rides in Halloween week. The bigger story is everything travelers ask AI before they book, and who writes the answers.

By Brouhaha Collective
AI booked four Uber rides in Halloween week

The AI agent that books your trip for you is still mostly a demo. Uber told analysts on its third-quarter earnings call that agents ordered four rides in the week of Halloween, its busiest week of October, out of roughly 320 million trips and orders across Uber and Uber Eats, TechCrunch reported. Agents placed about 500 food orders that week. DoorDash said it had received a few hundred agent orders in total.

The travel companies said much the same on their own calls. Airbnb's Brian Chesky told analysts that today's agents are good at "simple routine tasks like making a restaurant reservation," but "planning a trip is different," because "80% of travelers say they enjoy spending time researching trips" (TechCrunch). Expedia's chief executive, Ariane Gorin, said people still want to book with a trusted brand, "where they know that they'll be taken care of if something goes wrong" (PYMNTS). Expedia's own survey this spring found that only 8% of travelers feel comfortable booking through an AI platform.

It's easy to read all that as permission to relax. The robot travel agent was oversold, people still book for themselves, and a hotel or tourism office can sit this one out. The numbers say something narrower than that. They describe the last minute of a trip, the part where someone presses the button.

Fifty seconds in the Uber app

Uber's chief executive, Dara Khosrowshahi, offered the plainest explanation on the call: booking a ride in the app takes about 50 seconds, and reordering a favorite meal about 90. Nobody needs a robot to save them 50 seconds. Handing a credit card to software is a bigger ask, and in the same Expedia survey 66% of travelers said they wouldn't trust AI to make a purchase on their behalf. So people do the checkout themselves, the fast way, through the app or the brand they already use. None of that says much about where the decision got made.

The rest of Expedia's survey, as summarized by AltexSoft, does. Fifty-three percent of travelers were comfortable with AI suggesting travel options, 40% would use it to build an itinerary, and nearly half said it helps them find places they might not have found otherwise. Gorin put it in one line on the earnings call: "agentic AI's real strength is in discovery." On the same call, the company that calls its agent traffic too small to draw conclusions from also pointed out that third-party data shows Expedia leading in AI citation share in the U.S. Companies don't brag about numbers they think are irrelevant.

Who writes the answer

Discovery is where a hotel or a town gets chosen, or never comes up at all, and AI answers are built mostly from things the hotel didn't write. Muck Rack's latest "What Is AI Reading?" report, drawing on 207 million citations from AI responses this year, found that paid and advertorial content made up 0.3% of what AI platforms cited in September. Brand-owned pages were about 15%. Journalism was 22%, roughly seventy times what paid placement got, and most of the rest came from encyclopedias, government sites, academic research and other third-party content.

Two caveats, in plain terms. Muck Rack sells software for tracking exactly this, and its headline claim that more than 84% of citations are "earned" counts Wikipedia and the like, a definition the firm Fire on the Hill called too generous in PR Daily. Muck Rack's reply is fair, too: when it checked the Wikipedia pages of 489 S&P 500 companies, 60% of their references were journalism. And a PR agency telling you PR matters deserves a raised eyebrow of its own. Even on the strictest reading, though, about one AI citation in five is a news story, and almost none are ads.

The part travelers still do themselves

Picture a traveler asking an AI where to stay in a small coastal town in the off-season. The answer arrives built from travel stories, local guides, review sites and an encyclopedia page, nearly all of it written by people who don't work for the hotel. Then they book the place themselves, directly, because the card is already saved and it takes a minute. Every step before that minute is earned: the magazine piece that mentioned the new restaurant, the local writer's weekend guide, the town's page with the right opening dates. If those sources are good, the AI quotes them. If they're thin or three years stale, so is the answer.

The agents will get better. Airbnb plans to launch its own in the first half of 2027, and in November OpenAI bought Odessia, an AI travel agent that plans and books trips inside a chat, which sent Expedia, Tripadvisor and Booking Holdings shares down that day (PYMNTS). Investors, at least, aren't waiting for four rides to become five. But whenever software finally gets trusted with the credit card, it will still be choosing from what people have written about a place.

So stop treating AI as a booking channel to get ready for, and treat it as the newest reader of your coverage. The practical version is unglamorous PR: give travel writers a reason to come and something new to write about when they do, and keep the facts other people quote current wherever they live. Whoever tells a traveler where to go has done most of the selling before anyone opens an app, and for now that's mostly other people's writing.

Brouhaha Collective is an earned media agency for travel, tourism and the brands around them.

Work with us →
More from the Bureau