Heat waves (HWs) are natural hazards characterised by episodes of hot weather. However, in the absence of a universal definition a wide variety of definitions is applied. In this study, ten different air temperature (T) based HW definitions are applied to the urban region of Berlin, Germany, to investigate and compare the occurrence and duration of HWs, and their long‐term trends from 1893 to 2017. We studied how long‐term trends depend on different definition of HWs, as well as if long‐term mean values and trends differ between inner‐city and peripheral locations of Berlin. Generally, results show significant increases in HW occurrence and duration for most definitions, although large differences exist between them. Temporal agreement between the definitions is low, 15 episodes in 125 years are identified by all definitions as HWs. Inner‐city regions of Berlin are subject to more frequent and longer HWs than peripheral regions, if definitions based on daily minimum or mean T are applied. Results also show that trend estimations of HW characteristics for HW definitions with “extreme” values for their detection criteria (e.g., in terms of duration or threshold) are highly sensitive to the applied method. We conclude that depending on the question under investigation, different HW definitions might be optimal and hence attempts for the development of “universal” definitions need to take this into account.