
07/08/2025
Can LLMs Really Book a Complex Travel Itinerary? An Experiment.
We're constantly told that Large Language Models (LLMs) can create amazing travel itineraries, and will soon be able to actually book them for us. But how good are they really? I decided to put them to the test.
I asked ChatGPT, Gemini, and DeepSeek to perform a basic task for any experiential trip: create a table of 25 top events in Venice for June 2026 and provide booking links to relevant sites.
The results were telling:
* None of the LLMs could complete the task even after multiple attempts and various prompt changes.
* They all struggled with accuracy, using historic data for future events, and only one could even suggest more than 5 events.
* Website links were all generic, broken, or completely irrelevant (e.g. several links to "swimming with dolphins in Negril").
* They all issued disclaimers that any information provided needed to be independently verified before being used.
For an LLM to truly create and book a complex, event-based itinerary, it needs to be seamlessly integrated with a booking engine that has real-time inventory access and API permissions to tens of thousands of third-party activities, events, ticketing and experiences websites. But currently there’s no such platform - few of the aggregator platforms have significant real time inventory links to individual (rather than group) activities.
Our approach to experiential travel at Awesome Experiences is to use AI specifically designed for this particular task, so you don't have to wait for a paradigm shift in how the internet works to be able to plan and book your perfect trip. You can do it right now with technology available today, not in several years’ time.
Want to learn how we do it? Contact us at [email protected] or visit our website at www.awesomeexperiences.com.
; ; ; ; ; ; ; ; ; ; ; ; ;