Ask ChatGPT, Gemini or Claude to plan your next trip and the answers feel remarkably personal. Families get kid-friendly activity lists, budget travellers get cheap flight tips, and the tone shifts to match whoever is doing the asking. But a new Skift Research report suggests the personal touch mostly runs skin-deep: change the traveller, and the destinations themselves barely move.
The study, titled The Mystery of the Disappearing Destination: How AI Is Transforming Travel Discovery, put three major large language models through repeated test prompts using a range of traveller personas. These personas varied by demographics, interests, who they were travelling with and how much they wanted to spend. The models adapted their advice in the ways you'd expect. Personas travelling with children were pointed towards family-friendly activities, and those on tighter budgets were pitched low-cost flight options.
Where the personalisation fell short was the actual destination list. Whether the fictional traveller was a retiree, a backpacker or a parent of three, the same handful of well-known places kept coming back. According to the report, AI recommendations are almost as concentrated as real-world international arrivals, which cluster heavily around a small number of famous destinations.
That's a useful reality check for anyone using AI to plan a holiday. A chatbot that 'understands' you in conversation can still funnel you towards the same Paris-Rome-Bali shortlist it gives everyone else. If you want genuinely off-the-beaten-path suggestions, you may need to push harder — naming a region, excluding major cities, or asking specifically for lesser-known alternatives — rather than expecting the model to surprise you.
For travellers who already know where they want to go, AI remains a strong tool: itineraries, activity shortlists and budget breakdowns all benefited from persona-based tailoring in the tests. The gap is in discovery. If your goal is finding somewhere new rather than organising somewhere familiar, treat the chatbot's first answer as a starting point, not the full picture. The destination recommendations are closer to a popularity contest than a personalised match — at least for now.
