Why Can AI Replace Hotel Booking but Not Food Delivery?

Deep thoughts on AI and aspirations —— ByteDance Deep Thinking Circle

When Silicon Valley discusses AI entry points’ impact on platforms, the most widely circulated statement carries a hint of irony, called the “DoorDash Problem”: in the future, you won’t open a food delivery app, but instead tell an AI “order me a pizza,” and it handles everything, turning the platform into a pipe. But when you actually run the numbers, DoorDash itself is nearly the safest among all platforms, while hotel booking services are the ones truly standing at the cliff’s edge. Why do platforms in similar positions face such different fates?

Platform Profits Are Essentially an Annuity

First, let’s look at how platform businesses make money.

Food delivery, ride-hailing, hotel booking, e-commerce—all these platforms burn money on the same thing in the first phase: paying to acquire users. After acquisition, users develop the habit of opening the app directly, and every subsequent transaction costs almost nothing in customer acquisition. Trading upfront investment for long-term repeat purchases—that’s the platform annuity.

There’s another layer on top of the annuity: once users come on their own, ad placements and recommendation spots on the page can be monetized. Amazon’s advertising brings in about $60 billion annually, contributing nearly all the profit from its e-commerce business, with over half of product display positions being paid promotions.

AI entry points threaten both layers. When users start asking AI instead of opening apps, the annuity breaks: platforms must now pay for traffic on every single transaction. More painful is the second layer—when transactions happen elsewhere, the ad placements on the page disappear along with the transactions.

So far, no platform has actually died because of this. What’s broken is the logic: all commercial designs that assumed “users will come directly” as a permanent premise now have a loosening foundation.

And this scene isn’t playing out for the first time. Search engines already demonstrated the first half: Google did flight, shopping, and local search, directly moving the value of “helping you find” into its own results page, stripping away the profits of platforms that lived on referral traffic layer by layer. Those that survived were the ones that kept transactions and fulfillment in their own hands. AI entry points add one thing beyond search engines: they can not only find but also place orders for you. It’s completing the second half of a replacement that only went halfway back then.

The Dividing Line: Are You Selling “Finding” or “Backstopping”?

So why are hotel booking services the first to break?

Break down a platform’s value into four pieces: helping you find (search, price comparison), helping you transact (ordering, payment), helping you backstop (refunds, guarantees, financial terms), and helping you fulfill (delivery, logistics, service delivery).

AI entry points have completely different capabilities for these four pieces. Finding is what they’re best at—understanding intent, multi-turn questioning, cross-source price comparison—much stronger than platforms’ built-in search boxes. Transacting is being filled in, with instant checkout capabilities already rolling out. Backstopping and fulfillment show no signs of AI taking them on. These two mean building your own logistics, hiring drivers, tying up capital, handling disputes—all heavy-asset, low-margin grunt work that’s the exact opposite of model companies’ light-asset, high-margin business model.

So to judge a platform’s safety, don’t look at industry heat or technical narratives—just look at one thing: in its value proposition, how much is information and transactions versus responsibility and fulfillment?

Two variables amplify or reduce risk: usage frequency and decision weight. Ride-hailing and food delivery are high-frequency, low-decision-weight activities where users open the app out of habit without thinking—AI can’t wedge in. Vacations, buying ski equipment—these low-frequency, high-decision-weight purchases where users naturally research and compare extensively are exactly AI’s home turf. Even if you’re a loyal user of a particular platform, when booking a long vacation, you’ll most likely ask AI first.

The Counterintuitive Ranking: The More Thoroughly Digitized, The First to Fall

Applying the framework above to reality yields a ranking opposite to intuition.

The hotel industry is one of the earliest and most thoroughly internetized industries. Product standardization, data structuring, real-time inventory synchronization—all completed twenty years ago. Back then, this was a huge competitive advantage; now it’s precisely become a weakness: the more standardized and readable the supply, the easier it is for AI to aggregate. When model companies want suppliers to open accounts and integrate new channels, hotels have ready-made key account teams to serve them, and both sides hit it off immediately.

Contrast this with food delivery and ride-hailing supply: millions of local restaurants, fragmented vehicle capacity, and someone has to manage each order to ensure it’s delivered, on time, hot, and refunded if there’s a problem. Aggregating this kind of supply can’t be done with APIs alone. Short-term rentals fall between the two—more dangerous than hotels, safer than food delivery. Aggregating millions of rooms, each with its own layout and rules, took Airbnb many years and billions of dollars—photos, facilities, quality all had to be made readable one by one. This kind of grunt work can’t be copied by AI in the short term, but it can’t defend against scenarios beyond high frequency: travel planning itself is a low-frequency, high-consideration purchase, and traffic on the entry point side will still leak out bit by bit.

In other words, industries that digitized most thoroughly in the last era have the fewest defenses to give up in this era. The standardization painstakingly built back then has now become a manual everyone can read.

Hidden here is a second layer of blowback, which I call the advertising tax reckoning. Mature platforms have almost all been taxing user experience: more and more promoted positions stuffed into search results, rankings tilted toward higher bidders, commissions raised continuously. When there was no alternative, users endured it. Now that AI entry points have appeared, the first gift they hand users is a recommendation without ads. Every cent of tax platforms collected over the past twenty years has become a reason for users to leave. This is no longer a moral issue—the competitive structure has changed: for the first time, the portion of profit gained by degrading experience has a clearly priced competitor.

Four Ways for Platforms to Survive, and One Boundary

If you’re operating a platform or looking at platform investment targets, the response sequence is actually quite clear.

Do what AI won’t do. Deepen backstopping and fulfillment: insure transactions, build your own delivery, sign exclusive supply. These activities have low margins and slow results, but precisely because of this, model companies won’t follow. This is the only place you can’t be moved.

Search experience must stay close. You don’t need to lead—keeping the gap within twenty or thirty percent is enough; make up the rest with delivery and after-sales. If you’re really left ten times behind, users will leave completely.

Those with high market share should dare to stall. Doing e-commerce entry points can’t bypass the top platform—it holds inventory and fulfillment networks and is the needed party. At the negotiation table, slow is leverage; rushing to integrate is lowering your own value. If there’s real cooperation, data and decision rights must stay in your own hands.

The last one: before AI makes a move, proactively abolish the advertising tax. Passively waiting for others to disrupt you versus proactively cutting monetization that degrades experience—the two outcomes differ greatly.

The boundary also needs to be stated clearly. This framework has one premise: model companies maintain their light-asset DNA, only doing the understanding and decision layers, not touching fulfillment. If one day a model giant really pours heavy capital into building a logistics network, this safety ranking will need to be redrawn. But given their current business models, that’s unlikely.

Back to the opening question. Next time you look at a platform business, just ask one question: is it selling “helping you find” or “helping you backstop”? Those selling finding—the lease is up; those selling backstopping—they can still collect rent for many years.