Stravl: The world’s first large-scale, AI-based Travel Designer

Abstract

89% of Americans view travel planning as a stressful and time-consuming process (Turner, 2018) and with a growth in travel content available online, prospective travelers are expected to continue spending more time researching and planning their vacations. Stravl (deployed at Stravl.com) aims to counteract that trend by simplifying and personalizing the trip planning process using generative artificial intelligence and advanced machine learning algorithms wrapped in an engaging and gamified user interface. Stravl uses a custom-built extensive data set of user travel preferences as well as sophisticated language learning models (LLMs) to create the first large-scale, AI-driven Travel Recommendation System (TRS), providing users with unique destination recommendations and detailed trip itineraries customized to their preferences. Additionally, Stravl strives to provide users with a seamless, holistic travel planning experience with a wide range of secondary features, including group vacation dashboards, account systems, and destination exploration pages. The website’s primary objectives include optimizing the TRS for user engagement, long-term visitor retention rates, as well as general user satisfaction. Stravl aims to achieve that through deliberate website design choices that enhance the user experience and A/B tests that validate used methodologies. Since its launch, Stravl has gained 86,720 unique users and over 362,000 page visits originating from 175 countries. To the best of our knowledge, this makes Stravl the world’s largest AI-based TRS at the time of writing. This paper elaborates on the steps taken to build the TRS, including an in-depth view of its web-development and machine learning choices. Additionally, it evaluates user behavior and engagement analytics from wide-ranging A/B tests as well as direct user feedback.