Skift Report Shows AI Travel Advice Lacks Personalization

Artificial intelligence is growing more adept at sounding like it truly understands individual travelers. In a recent report titled The Mystery of the Disappearing Destination: How AI Is Transforming Travel Discovery, Skift Research tested three major large language models—ChatGPT, Gemini, and Claude—using repeated prompts across various traveler personas featuring distinct demographics, interests, travel companions, and budgets.

The models successfully adapted their approach; for instance, users traveling with children received details regarding family-friendly activities, while budget-conscious travelers were pitched low-cost flights. However, the actual destinations recommended by the AI models remained largely unchanged regardless of these varying personas.

The findings indicate that AI recommendations are almost as concentrated as global international arrival patterns, which consistently cluster around a handful of well-known and established hotspots. The top five destinations captured 38% of all AI recommendations, closely mirroring the 40% share those same five locations hold of actual global inbound arrivals. At the top 10 level, AI recommendations stood at 48% compared to 56% of real arrivals.

Stock photo for illustration only, not from the actual event

Even though destination variety remains narrow, AI still filters information in a way traditional search engines do not. While a search results page yields hundreds of links, an AI response provides a handful of options accompanied by specific reasoning. According to a joint survey by Skift Research and McKinsey involving 1,015 U.S. travelers, 63% preferred a curated shortlist of 10 options over a 500-option catalog at 30% or a single editable AI itinerary at just 7%.