Hydrogeochemical modelling has become an important tool for the exploration and optimisation of deep hydrogeothermal energy resources. However, well-established software and model concepts are often run outside of its temperature, pressure, and salinity ranges. The core of the model codes are thermodynamic databases which contain a more or less complete subset of mineral phases, dissolved species, and gases occurring in geothermal systems. Although very carefully compiled, they should not be taken for granted at extreme conditions but checked prior to application. This study compares the performance of four commonly used geochemical databases, phreeqc.dat, llnl.dat, pitzer.dat, and slop16.dat. The parameter files phreeqc.dat, pitzer.dat, and llnl.dat are distributed with PHREEQC and implement extended Debye–Hückel and Pitzer equations. Thermodynamic data in slop16.dat represent an implementation of the revised Helgeson–Kirkham–Flowers formalism and were converted to be used with the Gibbs Energy Minimizer ChemApp for this work. Test calculations focussed on the solubility for some of the most common scale-forming mineral phases (barite, celestite, calcite, siderite, and dolomite). The databases provided with PHREEQC performed comparatively well, as long as the range of validity was respected for which the corresponding database was developed. While it is obvious to use pitzer.dat at high salinities instead of phreeqc.dat, the range of validity for high temperature is usually not given in the database and has to be checked manually. The results also revealed outdated equilibrium constants for celestite in slop16.dat and llnl.dat and the lack of adequate temperature functions for the solubility constants of siderite and dolomite in all databases. For those two minerals, new thermodynamic data are available. There is, however, a lack of experimental thermodynamic data for scale-forming minerals in low-enthalpy geothermal settings, which has to be addressed for successful modelling.