Your website now has two visitors: what WebMCP means for travel brands

For the last two years, the question every travel brand asked about AI was how to get into the chat. Be the airline that answers in ChatGPT. Be the hotel Claude recommends.

This summer the industry started answering the opposite question too: what happens when the AI comes to your website?

The bet

On 25 August, OpenAI added WebMCP to the browser built into the ChatGPT desktop app. In OpenAI's words, a website can now "offer useful actions directly to an AI agent alongside the interface people already use." Google opened a public WebMCP trial in Chrome in May, and Microsoft runs one in Edge. The standard itself is being written in a W3C community group.

WebMCP is still early. It works today in Chrome and Edge trials and in ChatGPT's desktop browser. But when the largest assistant and the largest browser move the same way, it is worth understanding before your competitors do.

The interesting part is not the technology. It is the bet. OpenAI already lets brands bring their booking flow into ChatGPT as an app. With WebMCP it is also bringing the agent to the brand's own website, and its documentation says a website can support both. It is not choosing between the chat and the website. It is betting that travel will be booked in both, often in the same trip.

What it looks like for a traveler

A traveler is planning a long weekend in Rome. She starts in ChatGPT: where to stay, which neighborhood, what it might cost. ChatGPT suggests three hotels, including yours.

Today, the next step is the weak point. She clicks through to your site and starts over: destination, dates, two adults, one child, breakfast or not. Many travelers stop here.

With WebMCP, she opens your site with the agent still beside her. She says, "Book the family room for 14 to 17 March, breakfast included, and use my member rate." The agent works on your site, through actions your site has made available: it searches your availability, picks the room, applies her member price because she is signed in to her own account with you, and takes her to your checkout. She sees every step happen on your page, and she confirms the payment herself.

Two weeks later she asks the same agent to move the dates. It opens your site, finds her booking, and offers the change under your rules.

Nothing in that story happened on an online travel agency.

Why it matters for the business

A second chance at the direct booking. Travel brands have spent a decade pulling bookings back from intermediaries, and in AI chat the intermediaries moved first. WebMCP lets the brand's own website win the last step: the agent completes the booking where your prices, your rules and your margins apply.

Your member benefits travel with the guest. Because the agent works inside the traveler's own signed-in session on your site, member rates, points and saved details apply automatically. An intermediary cannot offer that, and it gives the agent a reason to finish the booking with you.

Less drop-off between interest and booking. The step where travelers abandon most often, re-entering everything they already said, disappears. The agent carries the request from the conversation to your booking engine.

Your website keeps its job. Many expected AI to make brand websites irrelevant. WebMCP points the other way: the website becomes the place where the booking is completed, verified and managed, with the agent doing the clicking.

What changes in the experience

Your site gets a second visitor. A person reads your pages. An agent reads the list of actions your site offers: search rooms, check availability, hold a fare, change a booking. If those actions are missing or vague, the agent falls back to guessing its way through your interface, and guessing loses bookings.

The page becomes the place where travelers check the agent's work. When the agent picks a room, the traveler should see it selected. When it changes the dates, the price should change on screen. A booking flow that only updates after a reload will look broken when an agent drives it.

Confirmation is a feature. ChatGPT asks the traveler to confirm purchases, account changes and messages. Design that moment as a clear, calm "review and book," not an interruption.

The banner stops selling. A hero image persuades a person. An agent reads what your site returns: room type, price, what is included, cancellation terms. If your reasons to book direct live only in banners, the agent will not pass them on.

What to do now

1. Choose the five actions that matter. For a hotel: search availability, show room options, apply member rates, start checkout, change a booking. For an airline: search fares, hold a fare, add bags and seats, change a flight.
2. Make member benefits part of the action, not a banner after login.
3. Make every agent action visible on the page, so the traveler can follow and trust it.
4. Keep the traveler in charge of payment, with a clear confirmation step.
5. Measure agent visits on their own. Track which actions agents use and how many end in a booking.
6. Do not rebuild your site. Start with the booking flow you already have, and keep it working for travelers whose browser does not support WebMCP yet.

Where this leaves travel brands

We already run this on our own site: bazak.ai offers a WebMCP action that lets an agent run our brand readiness scan, and we track every call. For our customers, Bazak runs the agent that sells inside ChatGPT and Claude and the agent that closes the booking on the brand's own site.

WebMCP adds the third case: the traveler's own agent arriving at your door. The travel brands that welcome it, with clear actions, their own prices and their own loyalty benefits, win the booking back from the intermediaries. The ones that do not will watch the agent book somewhere else.