The hotel industry has been here twice before. Both times, a new distribution layer appeared, hotels were slow to take it seriously, and by the time the response coordinated, the new layer had locked in 15–25% of every booking it touched. Forever.

The third reset is happening right now. The same mistakes are available again.

The two resets that defined modern hotel distribution

To understand what AI distribution actually is, look at the two prior shifts. The pattern is identical each time.

Reset 1: the GDS era (1978–2000)

When Sabre, Amadeus, and Galileo extended their airline reservation systems to hotels, properties that connected early showed up in every travel agent's terminal. Properties that waited became invisible to the agency channel — which, at peak, drove the majority of business travel.

The result: a permanent 8–15% take-rate on a vast slice of inventory, baked into every booking forever. Hotels that signed late never recovered the lost share.

Reset 2: the OTA era (1996–2015)

Expedia, Booking.com, and the others built a better consumer interface than any hotel could. Travellers comparison-shopped on the OTA, then booked on the OTA. Hotels that resisted lost ranking. Hotels that joined lost margin — typically 15–25% per booking, plus the guest relationship.

The OTA reset compounded over time. Each generation of traveller that came of age inside the OTA habit assumed it was how travel worked. The dependency became structural, and direct-booking campaigns have been clawing back single percentage points ever since.

The pattern

Both resets followed the same arc: new interface appears → hotels underestimate it → early adopters get disproportionate visibility → late arrivals concede margin to participate → the take-rate becomes permanent.

The decision window in both cases was roughly 3–5 years. The cost of waiting was paid for the next 20.

The third reset is now

The booking interface is shifting again. Not to a new website. To no website.

A traveller who asks ChatGPT, Claude, or Google AI "find me a hotel in Lisbon for the first weekend of June, under €250, walking distance to a beach" no longer needs to open ten browser tabs. The AI answers. Soon — and in some flows, already — the AI books.

This is the third reset, and it has three layers:

  • The discovery layer. AI assistants are increasingly the place travellers ask first. The result is a recommendation, not a list of links.
  • The recommendation layer. AI ranks and presents options based on the data it can ingest. Hotels that publish structured, machine-readable inventory get accurately represented. Hotels that don't get hallucinated, omitted, or misquoted.
  • The transaction layer. Agentic AI — assistants that don't just chat but actually execute — completes the booking. The protocol that lets the AI book directly with a hotel rather than an intermediary is being standardised right now under the name MCP.

The hotel that owns all three layers — visible in the recommendation, accurately represented, directly bookable — captures the guest with no commission. The hotel that owns none of them sees the AI default to whoever does: typically the OTA that was most aggressive about feeding the model.

What's structurally different this time

The temptation is to wait. AI is still maturing, the protocols are unsettled, real booking volume through agentic AI is small. Why move now?

Three reasons it's different from the prior resets:

Defaults form faster

The previous resets gave hotels years to react because the new interfaces took time to reach scale. AI assistants are scaling on rails the platforms already own — hundreds of millions of users, baked into operating systems, browsers, and productivity tools. When OpenAI or Google flip a switch on travel-booking workflows, distribution shifts in weeks, not years.

The data layer is the moat

In the GDS and OTA resets, the moat was the interface. In the AI reset, the moat is whoever's data the model trusts. That trust is being established now — through MCP endpoints, structured data, verified inventory feeds. The hotels indexed first become the model's reference answer for their market.

There is no consumer interface to compete with

In the OTA era, hotels could at least try to win the booking on their own website. In the AI era, the conversation is happening inside an interface the hotel doesn't control. The only lever is what the model can see and verify about your property. If the model can't verify, the model defers — usually to whoever already structured the data: the OTA.

"The question is not whether AI will mediate bookings. The question is whether hotels will own that connection or outsource it to intermediaries who extract rent." — Sigtrip Strategic Analysis, 2026

The window

Estimating the window is inherently imprecise, but the inputs are visible. MCP and equivalent protocols are stabilising through 2026. Major AI platforms have publicly committed to native booking workflows in the same window. OTA AI assistants — Booking.com's Trip Planner, Expedia's Romie, Kayak's GPT integration — are already shipping, optimised for OTA-favourable defaults.

Realistically, the strategic window for establishing direct AI presence is 12–18 months. After that, defaults harden. The hotels indexed first become the canonical answers. The hotels that arrive late will pay an OTA or an intermediary to be visible — exactly as in the prior resets.

The four moves for the next 12 months

  1. Audit your current AI visibility. What do ChatGPT, Perplexity, Google AI, and Claude actually say about your hotel today? You almost certainly don't know, and the gap between what they say and what's true is the work to do.
  2. Consolidate into a single source of truth. Rates, availability, policies, amenities, fine-grained operational facts. Today this lives in five systems. The AI layer needs it in one structured place.
  3. Publish a verified MCP endpoint. The well-known location at /.well-known/mcp.json is what AI platforms look for. Without it, the model can only scrape what third parties say about you.
  4. Treat AI distribution as a channel, not a marketing project. Assign ownership. Track visibility, mention rate, and direct AI bookings the same way you track OTA performance.

The imperative

Every distribution reset rewards the operators who treat it as infrastructure rather than novelty. The GDS-era winners weren't the hotels with the best terminals — they were the ones who treated GDS connectivity as table stakes early. The OTA-era winners weren't the hotels with the prettiest websites — they were the ones who built the operational discipline to manage rate parity, content, and channel mix.

The AI-era winners will be the hotels that treat AI visibility, structured data, and direct AI connectivity as the new table stakes — not next year, this year.

The third reset is here. The window is short. Your move.