Are major hotel chains rejecting AI agents? Test results have come out contradictory

Marriott, IHG, Accor and Choice hotel groups' websites let AI agents through to the payment stage in three consecutive tests. This result is presented in the second part of an independent study conducted with the participation of Claude, ChatGPT and Gemini tools. In the same tests, Jin Jiang and H World kept reservations behind a mobile-number login or app requirement.

Some of the results varied between days. Wyndham, which rejected Claude's availability query on 18 October, let the same agent through to the guest details page a day later. ChatGPT's Hilton error and the "no availability" result for Wyndham were not repeated in a re-test. Hyatt, meanwhile, blocked Claude on the same property where ChatGPT reached payment.

The study's main conclusion concerns who decides on access: no readable robots.txt file names an AI agent for the purpose of blocking it. Access is determined by bot management settings — Hyatt's block stems from Kasada's verification programme, whereas Choice, running the same programme, let both agents through. The authors emphasise that tests conducted over two days with two agents do not allow a definitive determination of whether the sector is open or closed to AI agents.