Every online travel agency now has an AI assistant. Booking.com shipped Trip Planner. Expedia rolled out Romie. Kayak integrated with ChatGPT. Tripadvisor launched its own. The product announcements landed within a 12-month window, and the framing was uniform: "personalised, conversational travel planning".

Read between the lines. These are not new products. They are a defensive perimeter.

What's actually shipped

Strip the marketing copy off and the assistants do roughly the same thing:

  • Conversational search over the OTA's existing inventory.
  • Trip itineraries composed from the OTA's hotel, flight, and activity catalogues.
  • Re-booking and modification through chat instead of forms.
  • Personalisation derived from the OTA's user history — the same data that drove their existing ranking and email-marketing engines.

None of this is technically novel. What's novel is the framing: the OTA is no longer presenting itself as a search engine for hotels. It's presenting itself as a travel concierge. The user is meant to bring the question and trust the answer, not comparison-shop the list.

The pattern: same play, different wrapper

The OTA business model has always depended on owning the moment of discovery. For 20 years, that moment was a search box on the OTA's website. The threat to that model isn't AI in the abstract — it's that the moment of discovery is moving into neutral AI assistants the OTA doesn't control.

If a traveller asks ChatGPT "find me a small hotel in Lisbon", and ChatGPT answers with three properties, the OTA may or may not be in the loop. The OTA has spent a decade building habit ("just go to Booking.com first"). Habit is exactly what neutral AI assistants threaten to overwrite.

The defensive logic

Every OTA AI assistant exists to re-anchor the discovery habit inside the OTA's own product. The bet is that "ask Booking.com's AI" becomes the new "go to Booking.com". The chat interface is the wrapper. The discovery monopoly is the prize.

What hotels actually lose

Three things to track, in increasing order of strategic seriousness:

1. The same commission, dressed up

A booking made through an OTA's AI assistant is, mechanically, the same booking made through the OTA. Same commission, same data ownership, same guest-relationship loss. The chat layer doesn't change any of the economics. It just hides them better.

2. Ranking opacity, multiplied

OTA ranking has always been a black box, but at least it was a ranked list — hotels could see roughly where they sat. An AI assistant presents one or two recommendations, often without showing the full set considered. The visibility tools that hotels rely on to monitor OTA placement no longer apply. "I'm on page one" becomes "I'm in the model's first suggestion" — and there's no way to check from the outside without testing prompts.

3. Discovery-surface lock-in

This is the one that matters in five years. If OTA AI assistants succeed at anchoring the discovery habit, then "asking an AI to find a hotel" becomes synonymous with "asking the OTA's AI". The hotel-direct opportunity inside neutral assistants like ChatGPT, Claude, and Google AI — which is the only AI channel that doesn't extract commission — never gets formed.

The counter-move

Hotels can't stop OTAs from shipping AI assistants. They can compete for the other AI surface — the one that doesn't charge them 18%.

The strategic priorities are unchanged from the broader AI distribution shift, but the urgency steps up:

  1. Audit your visibility in neutral AI assistants. What do ChatGPT, Claude, Google AI, and Perplexity say about your hotel today — without the OTA in the loop? The honest answer is usually "less than you'd hope, and often wrong".
  2. Publish machine-readable inventory directly. A verified MCP endpoint at /.well-known/mcp.json lets neutral AI platforms quote your live rates and availability without having to scrape an OTA listing.
  3. Make hotel-direct AI bookings frictionless. The moment a neutral AI assistant tries to book on the user's behalf, the path of least resistance wins. If yours has a verified endpoint and the OTA also does, the AI's default may go either way — but if only the OTA does, the default is decided.
  4. Track the OTA AI assistants as a separate channel. They will be measured by your OTA reps as part of OTA performance. They aren't — they're a new distribution surface with different ranking dynamics. Treat them accordingly.
"Whichever interface owns the discovery habit owns the margin. The OTAs know this. That's why every one of them shipped an AI assistant in 12 months." — Sigtrip Strategic Analysis, 2026

What to watch in 2026

  • Cross-platform integrations. OTA AI assistants embedded inside ChatGPT, Claude, or Google AI as MCP servers or plugins. This is the OTAs trying to be the canonical hotel source inside the neutral assistants too.
  • Rate-parity enforcement via AI. Watch for OTAs flagging properties whose hotel-direct AI bookings undercut OTA-AI rates. The next rate-parity war will be fought through the chat layer.
  • "AI exclusive" inventory. Promotional rates surfaced only through the OTA's assistant, locking users further into the OTA habit. Likely first in flights, then hotels.

The OTAs are racing to be the AI travel default. The window for hotels to establish a credible alternative — direct presence inside neutral AI assistants, backed by verified data — is the same 12–18 months that defines the broader distribution reset.

The chat wrapper changes. The economics don't. Plan accordingly.