Booking platforms do real work. They aggregate availability, standardise listings, process payments, collect reviews, handle cancellations, reduce fraud, and support travellers when something goes wrong.
That infrastructure is valuable. It is also bundled.
When one platform owns discovery, ranking, identity, payment, messaging, and dispute operations, the fee pays for more than payment processing. It pays for access to the platform's complete coordination system.
The interesting question is not why booking platforms charge money. It is what happens when agents can perform part of the coordination directly, while the participants buy verification, payment, insurance, and dispute services only when they need them.
Platforms bundle the transaction
Consider a four-night hotel reservation. A traveller searches, filters, compares rooms and policies, enters details, pays, receives confirmation, and relies on the platform if the booking fails. The hotel provides the room; the platform provides the environment in which the hotel and traveller can discover and trust one another.
That environment has historically been difficult to assemble. A hotel needs distribution. A traveller needs confidence that the room exists, the price is real, and the cancellation terms will be honoured. The platform solves both problems at once and charges for the bundle.
The resulting fees are material. Airbnb currently documents split-fee stays in which most hosts pay 3% and guests generally pay 14.1–16.5%, and a single-fee structure in which most hosts pay 15.5% while others typically pay 14–16%. Booking.com does not publish one universal accommodation commission: its terms calculate commission using the percentage established in the accommodation agreement and explain that commission can affect ranking. These structures vary, and they can change. The point is not one exact percentage. It is that control of coordination carries economic value.
Now give both sides an agent
Imagine that a hotel agent knows current room availability, acceptable prices, check-in constraints, cancellation rules, breakfast options, accessibility details, and which exceptions require human approval.
The traveller's agent knows the destination, dates, budget, party size, arrival time, dietary needs, room preference, cancellation requirements, and which compromises are acceptable.
The traveller's agent can publish or send a structured request:
Four nights in Bologna, two adults and one child, arrival after 21:00, cot required, vegetarian breakfast preferred, budget below €900, flexible cancellation until seven days before arrival.
A hotel agent can evaluate the complete request against current operating context and reply with a specific offer: price, room, included services, cancellation terms, tax treatment, check-in confirmation, and expiry time.
The traveller's agent can compare offers, ask follow-up questions, and present the important differences. The person still approves the consequential decision. The agents remove the repetitive search and clarification around it.
This is not automated website browsing
An agent that opens a marketplace, clicks filters, and copies a confirmation number has automated the old interface. It has not changed the market structure.
The structural change begins when customer and business agents can discover one another through shared, public records and exchange context through a common protocol. Neither side needs to predict every possible journey as a fixed series of pages and forms. They can start with a compact record and reveal deeper information only when the match is credible.
This is progressive disclosure applied to a transaction. First: a room may be available in the right place and time. Next: the room, policy, and price fit the request. Finally: the agents assemble explicit terms for human approval.
Detailed coordination becomes cheaper because agents can ask the annoying questions without consuming constant human attention: parking, late arrival, allergies, cots, stairs, extensions, deposits, or changes in room type.
Prices can respond to the transaction
A marketplace needs rates that can be displayed and ranked at scale. An agent can consider the actual conditions of a transaction.
A hotel may accept a different offer when occupancy is low, dates are flexible, the stay is longer, cleaning requirements change, or an external distribution fee can be avoided. That should not mean opaque pricing based on personal identity. It should mean an explicit offer based on transaction-relevant conditions: timing, availability, flexibility, service level, cost, and risk.
Suppose a hotel wants to receive €850. Under an illustrative 15% distribution fee, it would need a €1,000 booking to preserve that amount. If direct coordination, payment, and verification instead cost €25, the hotel could offer €900, receive €875 before operating costs, and leave the traveller €100 better off.
Those numbers are an example, not STAP pricing. They show the structural possibility: when coordination becomes cheaper, both sides can benefit without either capturing the entire saving.
Trust is the difficult part
Cheap communication alone produces spam, false availability, fake properties, payment fraud, and promises nobody intends to honour. Removing the marketplace while removing its protections would not be progress.
The alternative is modular trust. An agent should be able to inspect persistent identity, source provenance, accuracy history, fulfilled terms, disputes, current evidence, and accountable economic commitments. Payment, insurance, verification, and resolution can still be supplied by specialised providers.
STAP's Committed Trust Certificates are one part of that system. CTC is platform collateral and service credit, not insurance or payment to the counterparty. Signed transaction terms can reserve CTC and define a liability amount. If a manually reviewed dispute creates a valid claim, the deduction is limited by the remaining signed liability. The winner does not receive the deducted CTC.
The mechanism does not guarantee a good stay. It makes the commitment explicit, attributable, and consequential.
From one platform to modular services
Closed platforms still have major advantages: liquidity, familiar interfaces, standardisation, fraud operations, customer support, and regulatory expertise. Agents do not make those functions disappear.
They do make the bundle contestable.
One provider can supply discovery. Another can verify the hotel. Another can process payment or insure the stay. Another can resolve a dispute. The hotel and traveller can retain a direct relationship while paying for the infrastructure they actually use.
Booking is only one example. The same structure applies to local quotes, spare transport capacity, restaurant orders, equipment rental, and other markets where a large fee pays for the difficulty of finding, evaluating, and coordinating with the other side.
The goal is not zero infrastructure and not zero cost. It is to move the fee toward the cost of coordination, verification, payment, and risk—instead of the value of controlling the only index where the participants can meet.
The customer's agent should be able to ask the real world. The hotel's agent should be able to answer. Clear terms and accountable trust should let the transaction proceed without requiring one company to own the entire relationship.
Fee information checked August 7, 2026: Airbnb service fees, Booking.com General Delivery Terms, and Booking.com: How we work.