Wyndham Hotels and Resorts has placed its own agents inside ChatGPT, Claude and Gemini, supplying the three assistants with real-time rates, inventory, guaranteed room types and personalised loyalty perks, Skift reported on Thursday. The report previews the company’s session at Skift Global Forum in New York next week and sets out Wyndham’s wider AI strategy for search and distribution.

The move reverses the usual order. Most hotel groups wait for the models to find them, live with whatever the model has scraped, and hope the recommendation comes out right. Wyndham is feeding the models directly, so a traveller who asks an assistant for a room gets the group’s own rate and its own loyalty benefit rather than a third-party summary. Hotels, OTAs and the loyalty teams inside both are the ones this hits first. The approach amounts to a hotel AI distribution strategy that breaks into five decisions, each one small enough for a hotel group or OTA of any size to copy.

1. Put the agent where the traveller already is

Wyndham’s first step was a native app inside ChatGPT, launched on 6 May 2026. Users explore its hotels on a map, with amenity filters and hotel cards, without leaving the chat, then complete the booking on WyndhamHotels.com, the company said in its launch announcement. The agents now inside Claude and Gemini extend the same conversational AI hotel booking model. The assistant does the discovery and Wyndham keeps the transaction.

2. Point AI at the AI

The second step is the one most groups skip. Wyndham uses AI to monitor how travellers phrase long-form searches inside the assistants, then adjusts the data it supplies in response. After seeing searches for beach hotels with “dedicated beaches with umbrella coverage”, it added that detail to its feeds. The feed is treated as a living product, rewritten around the questions people actually ask, not a one-off export from the property system.

3. Give the agents something to sell

An agent with nothing to upsell is a brochure. More than 5,000 hotels are live on Wyndham Connect, which generated over USD 11 million in upsells in 2025. The Wyndham AI Concierge converts 35 per cent of calls into bookings. Franchisee retention stands at 96 per cent, which matters because every feed and every perk the agents offer depends on owners staying in the system.

4. Fund it over years, not quarters

None of this came from a single budget line. Wyndham has invested more than USD 450 million in technology since 2018. AI has cut its average call centre handle times by 7 per cent. The company says top-performing hotels averaged more than USD 60,000 in incremental revenue, with the highest above USD 200,000. Those are property-level numbers, which is how a franchisor persuades owners to keep paying for the platform year after year.

5. Partner rather than build everything

“Conversational AI is reshaping how travelers discover and book,” said Scott Strickland, chief commercial officer at Wyndham Hotels and Resorts, in the ChatGPT app announcement. The partners Wyndham names are the model companies themselves, OpenAI, Anthropic and Google, alongside Oracle, Amazon, Adobe, Salesforce, Canary and Aven Hospitality. The group’s own work sits in the feeds, the loyalty rules and the booking path. The models and the middleware are bought in.

Who to watch

Geoff Ballotti, president and CEO of Wyndham Hotels and Resorts, is a confirmed speaker at Skift Global Forum 2026 at the North Javits Center in New York on 22 to 24 September. His session is the place to hear Wyndham’s AI strategy laid out by the company itself, with the numbers behind the Claude and Gemini agents. Organisers planning an AI distribution panel for hotel owners have a ready-made case study in this one, and a franchise-heavy operator is the version of that story most owners in Japan, India and Singapore can act on.

Key takeaway

We read this as the first published AI distribution playbook that works below the luxury tier, and the cost of copying it is mostly discipline rather than money. Any hotel group or OTA with a loyalty programme and a live rate feed should test steps one and two this quarter: expose a rate feed to at least one assistant and start reading the long-form queries that come back. Groups without a direct booking path to hand the assistant to can wait, because an agent that passes the transaction to someone else is just a new commission line.

The post Wyndham’s 5-step AI playbook: own agents in ChatGPT, Claude and Gemini appeared first on TravHQ.