Two brands can look identical on AI visibility and be nowhere near each other when it actually matters

Two brands can look identical on AI visibility and be nowhere near each other when it actually matters.

I audited the major cruise lines to understand how AI models mention and cite each brand when someone asks about booking a cruise. I looked at both ungrounded (memory) and grounded (live retrieval).

The memory leaderboard looked normal: every brand showed up as expected. The real threats, though, weren't aggregators; they were rival cruise lines, the two most common names appearing in about half of all responses. Aggregators showed up too, just less often, in the 23–30% range.

But the picture flipped completely on grounded data. Two of the four saw their own Share of Model (SoM) roughly halved — one from 9.8% to 4%, another from 9% to 3.5% — while the category leader actually gained ground, from 13.7% to 16%. More telling: in grounded mode, all ten top entities were booking aggregators, some at 46% share. Not one rival cruise line made the list.

What travel brands should pay attention to: as more people lean into AI models, those models route them to aggregators, not the brand. The industry built its distribution rails — OTAs, GDS — decades before AI and aggregators now own the live infrastructure agents need to quote real-time availability.

That creates a margin problem: an OTA booking carries 10–15%+ commission you'd never pay on a direct sale. On top of that, you miss out on first-party data. No email, no loyalty sign-up, no way to win that customer back directly. So it doesn't only cost you margin once; it costs you the repeat-booking economics because the aggregator owns the relationship from there. It happens one AI answer at a time, invisible until someone measures it.