Groundwater databases are needed to systematically and uniformly compile the extensive and diverse data from previous surveys, providing it in its original electronic format. This paper categorizes the diversity of groundwater data classification and reviews the current status of domestic and international groundwater databases, including the “Regional Groundwater Information Database” of the Japanese Association of Groundwater Hydrology. It addresses long-discussed issues in groundwater data construction: spatiotemporal variability, data uncertainty, and scale dependency. Additionally, this paper discusses the requirements in an ideal groundwater database that serves academic and practical purposes. It also explores methods for equipping databases with generalization capabilities to estimate or predict variables at arbitrary locations without previous data, using three approaches: spatial interpolation, numerical simulation, and machine learning.