The use of automated driving systems (ADSs) in the real world requires rigorous testing to ensure safety. To increase trust, ADSs should be tested on a large set of diverse road scenarios. Literature suggests that if a vehicle is driven along a set of geometrically diverse roads—measured using various diversity measures (DMs)—it will react in a wide range of behaviours, thereby increasing the chances of observing failures (if any), or strengthening the confidence in its safety, if no failures are observed. To the best of our knowledge, however, this assumption has never been tested before, nor have road DMs been assessed for their properties. Objective/Aim. Our goal is to perform an exploratory study on 47 currently used and new, potentially promising road DMs. Specifically, our research questions look into the road DMs themselves, to analyse their properties (e.g. monotonicity, computation efficiency), and to test correlation between DMs. Furthermore, we look at the use of road DMs to investigate whether the assumption that diverse test suites of roads expose diverse driving behaviour holds. Method. Our empirical analysis relies on a state-of-the-art, opensource ADSs testing infrastructure and uses a data set containing over 97,000 individual road geometries and matching simulation data that were collected using two driving agents. By sampling random test suites of various sizes and measuring their roads’ geometric diversity, we study road DMs properties, the correlation between road DMs, and the correlation between road DMs and the observed behaviour.