Council Post: Why Hotel Distribution Is Still Broken, And What Comes Next
Abhinav Sinha is Co-founder and CEO of ZentrumHub, a unified hotel API platform serving 90+ OTAs globally.gettyA couple of Saturdays back, a supplier asked us to block a user: 3,000-plus searches on one propert...
Abhinav Sinha is Co-founder and CEO of ZentrumHub, a unified hotel API platform serving 90+ OTAs globally.
getty
A couple of Saturdays back, a supplier asked us to block a user: 3,000-plus searches on one property, one set of dates, in 40 minutes. Normally that's fraud. It wasn't. It was an AI agent, a model wired to a headless browser. Every query was real: flipping occupancy, swapping rate plans, re-shopping every five minutes. Two days of a travel agent's phone calls were done before the supplier's team finished breakfast.
The supplier thought it was abuse. I thought it was their future customer. On a Saturday, that argument wasn't winnable.
Here's the part nobody in our industry wants to put in writing: Hotel distribution isn't broken because it's fragmented. Fragmentation has been a symptom since 1995. The problem is older. Every layer of the stack assumes shopping is scarce and expensive.
The GDSs, the global distribution pipes that agents have booked through since the 80s, charge for every segment of a booking. Bedbanks, the wholesalers who buy rooms at net rates and resell them, price their discounts assuming you won't run a thousand searches per booking. OTAs take commissions. Parity clauses, the fine print stopping a hotel from selling cheaper anywhere else, held it all in place. A look had human cost. A look was supposed to convert.
Agents don't shop to convert. They shop to reason. That sentence is the whole piece.
In my company's rate-debugging dashboards, a hotel search routinely hops through seven, sometimes 14, systems before it reaches a screen. These systems include supplier, switch, aggregator, channel manager and the caches wedged between, because every layer upstream charges for live calls and every layer takes a margin. That's why your rate vanishes at checkout, why one hotel shows on 30 channels at 30 slightly different prices and why a nice five-star in Bali takes nine seconds to say whether a room is free.
Two numbers never make it onto panels. The mapping platforms this industry depends on advertise 99.99% accuracy across hundreds of suppliers. This is self-reported and never independently audited. The entire agentic travel economy is built on that one sentence. And HBX, public since its February 2025 IPO, reports a take rate of 8.8% on 8.2 billion euros of transaction value. This is the first public price on a bedbank's layer.
The previous fixes didn't work. Hotel NDC, an attempt to port the airlines' New Distribution Capability standard to hotels, was mostly a PDF with a nice logo. The 2015 to 2019 direct-booking push was partly theater; Kalibri Labs puts total direct acquisition cost at 15% to 25% of revenue once you count brand PPC and retargeting. Blockchain was a rounding error. All shared one premise: The stack underneath is sound and just needs a cleaner API on top.
Here's the bit I expect pushback on; if I'm wrong, I'd rather find out in writing than at a conference. The commercial DNA of this industry assumes the shopper is scarce. The shop itself is the expensive part.
Cloudflare's 2025 year-in-review found AI "user action" crawling (bots visiting sites because a person asked a chatbot) up more than 15 times in a year. HUMAN Security measured agentic traffic growth of 1,300% between January and August. Still no published 2025 hotel look-to-book benchmark from Triptease, HEDNA, RateGain or SiteMinder. The data vacuum is the story.
On parity: In September 2024, the EU's top court ruled that Booking.com's parity clauses can't hide behind the "ancillary restraints" shield in competition law, leaving them open to challenge. Booking had already dropped them across the EEA that July following its designation as a Digital Markets Act gatekeeper.
More than 15,000 hotels have lined up behind HOTREC, the European hotel association, to claim damages. HOTREC's last study put Booking at 71% of Europe's OTA market; the next edition, due this year, will show whether any of this moved the number. My bet: barely. Parity was never just a contract.
It was a ranking algorithm plus a Preferred Partner surcharge plus a Genius discount plus a Google Hotel Ads bid. The language got banned. The algorithm stayed. But agents don't read the sort. The sort is the ranked results page, the thing that machinery exists to influence. An agent reads a structured response, reasons over it, picks. The assumption that the shopper is a human looking at a screen is going away faster than the industry will admit.
The next layer isn't another aggregator. Aggregators earn per shop. Whatever comes next earns per decision, per booking, on inventory exposed in machine-readable form. Anthropic's Model Context Protocol, OpenAI and Stripe's Agentic Commerce Protocol, Google's Agent Payments Protocol and the card networks' agent programs all converge on the same primitive: one standard way for an agent to find inventory, prove it's authorized to buy and pay. The money sits on the booking, not the search. The alphabet soup is annoying. The economic inversion underneath is the story.
The marginal cost of a search heads to zero. The toll moves to conversion.
For 20 years, this industry optimized for ADR, average daily rate, defending it by defending scarcity: of inventory, of rates and, though nobody wrote it down, of search. The scarcity of search was real in 1987, when a query cost actual money to serve. It stopped being real decades ago when that cost collapsed. We just kept charging as if it hadn't.
Anyone whose unit economics depend on per-shop revenue is setting themselves up for failure. Metasearch cost-per-click stops making sense because agents don't click. They select. And rate parity, out of contracts for 18 months but kept alive through rankings and discounts, finally runs out of ways to be enforced.
The winners will be chains that can expose clean machine-readable inventory without getting scraped into oblivion, and infrastructure companies willing to solve the actually hard problem: translation. A traveler types "somewhere quiet for our anniversary, under four hundred a night, near water, not a brand I've stayed at before," and something has to turn that into a deterministic query across 40 supplier formats. Connectivity stops being the moat.
Therefore, the choice isn't whether agents should shop this way; they aren't going to ask our permission. It's whether your inventory is legible when your best customer searches on a Saturday morning.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?