This insight summarises Are Hotels Closing Their Doors to AI Agents?, by Terence Ronson, Founder and Managing Director of Pertlink Limited, published on Hospitality Net in October 2026.
The question. After reports of websites blocking AI agents, Pertlink tested whether agents can read information and reach checkout on the sites of eleven of the world's largest hotel groups. Claude and ChatGPT ran live browser tests on 18 October 2026 for one adult and one night in November. Gemini produced a viewpoint paper with no stated method. No bookings were submitted and no personal or payment details were entered.
Key findings:
- Most groups are not shut out. Seven of the eleven let at least one agent reach checkout. Marriott, IHG, Accor and Choice admitted both.
- Access is inconsistent. Hyatt stopped Claude with a bot challenge but let ChatGPT reach payment. Hilton let Claude through but returned a 502 error to ChatGPT. Wyndham showed neither agent a rate.
- China's major groups gate booking by design, using mobile-number logins, SMS codes or app handoffs.
- Policy rarely blocks AI. None of the readable robots.txt files named AI agents for blocking, and a separate study of 105,002 hotel websites found only 3.3% set any rule against AI crawlers.
- Security is the real gatekeeper. Malicious bots made up 44.5% of online travel traffic in Imperva's 2024 Bad Bot Report, so bot defences decide access in practice.
- AI research needs checking. Gemini reported every group as blocked at reservation, which both live tests contradicted.
"Sold out" and "blocked" can look identical. When Wyndham's availability request was dropped right after a bot-detection check, the agent, and the guest behind it, could not tell a silent refusal from a sold-out hotel.
Conclusions. Access varies by agent and route rather than hitting a clear wall. Published policy and actual enforcement are misaligned. Payment and identity remain the last hurdle, since no test completed a booking. Emerging standards such as signed agents, Google's AP2 and MCP were not observed on hotel group sites.
Why this matters for hotels. In Ronson's words, "hotels have not closed their doors to AI agents, but most have not decided how to open them either." Every hotel and group should test its own website for agent access, bot-defence precision and data accuracy, and align security settings with its distribution strategy.