Planning a trip is about to stop being a job

Somewhere on your computer there is a graveyard of trip research: a browser window with forty tabs, a spreadsheet of flight prices that expired an hour after you pasted them in, and a Google Map studded with pins for restaurants you never reached. For two decades, planning a trip has been a job, one we assigned ourselves and did at night after the actual job. I build travel software for a living (Sola, an iOS app for women who travel solo), so I watch this ritual from inside the industry, and I think it is in its final years. The forty tabs are dissolving into a conversation, and the conversation is starting to dissolve into a task you hand to software. The question worth thinking about is which parts of the job we should be glad to lose.

The conversational turn has a start date. Expedia and Kayak shipped the first travel plugins for ChatGPT in March 2023, Expedia added a ChatGPT planning beta inside its own app that April, and Booking.com followed with an AI trip planner in June. Expedia's chief technology officer, Rathi Murthy, said planning a trip becomes "simply starting a conversation." Those first plugins were clumsy, limited to paying ChatGPT subscribers, and later scrapped entirely, but the direction held. Google previewed a dedicated Gemini trip planner in May 2024, then shipped the idea as ambient features instead: since March 2025 its AI Overviews have built day-by-day itineraries for whole regions inside an ordinary search. Even Google's own planner dissolved into the conversation rather than becoming an app you open.

Then the industry started building for the handoff itself. In January 2025 OpenAI launched Operator, a research preview of an agent, meaning software that operates a website the way a person would, clicking and typing its way through to complete a booking. It launched for American subscribers paying 200 dollars a month, which tells you how early it was. In April the card networks moved a day apart: Mastercard announced Agent Pay on the 29th and Visa announced Intelligent Commerce on the 30th, payment rails built specifically so that software can buy things on your behalf. Visa's Jack Forestell put it plainly: "Soon people will have AI agents browse, select, purchase and manage on their behalf." By September Stripe and OpenAI had launched Instant Checkout inside ChatGPT, with an open protocol for agent purchases, Etsy sellers live first and over a million Shopify merchants promised as "coming soon." And in October Expedia and Booking.com became two of the first apps inside ChatGPT, so you can now search flights and hotels inside the chat itself, though when you go to book, ChatGPT hands you back to the travel site to pay.

Then reality did what it does to demos. By March 2026, Shopify told Forrester that roughly 30 merchants were actually live on that agent-checkout protocol, far short of the million-plus promised, and Skift ran the headline "ChatGPT Bails on Transactions." Discovery has genuinely moved into the conversation. The transaction, for now, still runs on the old rails.

Travellers moved anyway. Phocuswright found in March 2026 that 56 percent of American travellers had used AI for planning, booking or help mid-trip on at least one trip in the past year, up from 33 percent at the start of 2025, the fastest behavioural shift the firm has measured in a decade. The same study carries the caveat that matters: only 8 percent felt the AI's answer alone was enough, and 51 percent clicked through to check the sources. The tabs are shrinking rather than gone.

That checking instinct is still earning its keep. Last year two tourists in Peru paid nearly 160 dollars to be taken toward the "Sacred Canyon of Humantay," a destination an AI had invented by stitching together unrelated place names; the local guide who intercepted them told the BBC that this is how travellers end up at 4,000 metres with no oxygen and no phone signal. A chatbot writes a fluent itinerary for a place that does not exist with exactly the same confidence it brings to a place that does.

There is a second side to this shift, and it is the one I watch professionally: what it does to the places trying to be found. The forty tabs were a terrible interface for travellers, but they were an open market for destinations. A small guesthouse could earn one of those tabs with good photographs and ten years of honest reviews. When the recommendation comes out of a model instead, that market goes opaque. Hotels have noticed: by April 2026 the trade press was covering visibility tools for what the industry now calls generative engine optimisation, the craft of getting recommended by chatbots the way businesses once fought to rank on Google, and reporting that properties are already showing up in AI travel recommendations while "most have no idea what's being said about them or what's driving those results." The practical advice for places is unglamorous: describe yourself accurately and consistently everywhere machines read, and give guests and writers real things to say about you, because the models learn from what gets written down.

And then there is the part of the job I hope no one delegates. A Chase Travel survey found that 44 percent of travellers feel more overwhelmed planning trips than they used to, and that 60 percent of Gen Z and millennial travellers want to use a human travel advisor. Chase owns travel-advisor businesses, so treat those numbers as directional, but the overwhelm rings true, and note what people are reaching for: someone to carry the admin rather than someone to want things for them. Choosing is a different category of work. A 2010 study of 1,530 Dutch adults found that people taking holidays were happier than those who were not, but mostly before the trip; the researchers put it down to anticipation, which press coverage of the study reported kicking in as much as eight weeks before departure, and for most travellers there was no lasting happiness boost after coming home at all. Deciding where to go, and then reading and imagining your way toward the place for weeks, is a measurable share of what a trip actually gives you. Hand an agent the deciding and you have outsourced the happiest part.

So here is where I land, as someone building in this wave and cheering most of it. The job part dies and good riddance: the price refreshing and the tab reconciling and the late-night cross-referencing of a 2019 forum thread against a hotel map will go into the conversation, and the paying and rebooking will follow once the rails catch up with the announcements. What should never go is the agency. The reasons you travel, and the taste that says this town rather than that one, belong to you; so does the willingness to be wrong about a place in person. Any product that tries to automate the wanting has solved the wrong problem. Give the machine your admin, and keep the wanting for yourself.