This insight summarises OTAs Are Betting on Traveler Trust. But the Scramble Is On to Win the Trust of AI Agents by Dennis Schaal, published on Skift on 2 July 2026.
Two different trust bets are being placed in parallel. Expedia and Booking.com executives have staked their AI-era pitch on consumer trust — the argument that travellers, wary of ChatGPT hallucinations, will keep coming back to established travel brands to actually book. Trip.com Executive Chairman James Liang, on the same earnings-call cycle, framed the challenge differently: the trust that matters isn't consumer-facing at all — it's the trust the LLMs place in you when they decide whose inventory to surface.
"Trusted infrastructure for AI agents." Liang's phrase from the Trip.com earnings call captures the shift. Trip.com is positioning to power AI-native booking interfaces from the inside by packaging verified inventory, real-time pricing, and reliable transactional data into feeds the LLMs can consume with confidence. While Expedia and Booking work to keep travellers from defecting to AI-first booking flows, Trip.com is trying to be the plumbing underneath those flows.
Why this matters for suppliers. If a small number of OTAs win the "LLM's favourite data provider" contest, discovery consolidates around whoever the models decide to trust. Travellers may not care whether they trust an LLM — if the LLM only trusts a handful of aggregators for hotel inventory, the aggregator's grip on distribution tightens rather than loosens. That inverts the story most independents have been told about AI democratising discovery.
The competitive dynamics to watch:
- Exclusive vs. multi-partner LLM integrations — a single-OTA arrangement per model would concentrate distribution power dramatically.
- Model Context Protocol (MCP) adoption — the mechanism through which OTAs (and suppliers directly) feed structured inventory into AI planners.
- Verification and data quality — becoming the "trusted" source is a moat built on accuracy, not marketing.
- Direct-supplier MCP feeds — hotels and airlines that expose their own trustworthy MCP endpoints can bypass the OTA-as-infrastructure layer entirely.
The takeaway for hoteliers: the OTA distribution question in the AI era isn't just "will my rate parity hold?" — it's "which OTAs are the LLMs going to trust, and do I have my own path to trust with them?"