Tsunami databases contain general information of historical tsunamis, which provides valuable and fundamental data to tsunami-related researches. Nevertheless, before adopting the information stored in various tsunami databases, users should be familiar with these databases’ individual features due to their inhomogeneity in data storage. To this end, a comprehensive review of the worldwide existing tsunami databases was conducted in this study. Two kinds of existing databases are recognized, i.e. the individual database and the organizational database. The individual tsunami databases are generally simple and descriptive with less information about tsunami and associated earthquake. Update of individual database is also out of date. Nevertheless, the organizational tsunami databases, with the retrieval function and timely update, are more sophisticated with significant research orientated information, upon which we can obtain the detailed characteristics of a specified tsunami event. In addition, a further comparison is made between two main organizational databases, i.e. the databases of NTL/WLD and NGDC/WDS. In which, the number of recorded tsunami wave heights per event and the number of definite tsunami events per decade both witness an apparent increase after 1900s, and the landslide-generated tsunamis are found to gradually increase after 1960s. The total tsunami events per decade fluctuate around one hundred since 1990s, whereas the definite tsunami events per decade experience a linear increase in this period. It turns out that though quite a part of the data in NTL/WLD is referred to NGDC/WDS, they do collect different tsunami events, and even for the same event, their recording datasets are still not consistent with each other. Figuring out the structural features of different databases and unveiling their connections and differences facilitate relevant research for disaster prevention and mitigation.