Google's AI Max Can Send a Hotel's Ad Budget to a Competing Property

Google’s AI Max advertising feature can send a hotel’s own paid-search budget to a competitor’s front desk — and the mechanism is built into the product, not a bug. Kylie Chen, Director of Paid Strategy at Cogwheel Marketing & Analytics, flags two specific risks for independently branded hotels operating under a shared brand URL structure — a Hilton Garden Inn or Marriott property, for instance: broad-match search-term expansion that can pull in irrelevant queries like “1 800 marriott” or “marriott central reservation,” and automatic final-URL expansion, where Google “automatically sends users to the landing page from your domain that is predicted to perform best” — which can route a paid click to a different property in the same brand family rather than the one that paid for the ad. Chen’s conclusion is pointed: AI Max’s automation is a genuine advantage for large brand campaigns and OTAs advertising many properties at once, and a genuine liability for single-property operators trying to protect direct bookings.

The risk here isn’t that AI is bad at hotel marketing — it’s that AI optimized for a platform-wide objective can work against a single property’s objective even while technically “performing well” by Google’s own metric. That’s the same distinction Cendyn CMO Nicola Graham draws from the other side of the martech stack: most hotels aren’t short on data, they’re short on direction, drowning in dashboards without a clear sense of which signal deserves action. Graham frames AI’s real value as prioritization — surfacing the handful of actions that matter in a given moment rather than adding another automated layer for someone to manually audit. AI Max is what happens when that audit doesn’t happen: automation acting on an objective the hotel never explicitly chose.

For independent and flag-affiliated hotels alike, the actionable move isn’t avoiding AI-driven ad tools — it’s auditing what objective the automation is actually optimizing for before turning it loose on a live budget, especially inside a shared-brand URL structure where “predicted to perform best” and “best for this specific property” can quietly diverge.