In detecting the structure of a lithography mask, a portion of the lithography mask is firstly illuminated with illumination light of an at least partially coherent light source in the at least one preferred illumination direction. A diffraction image of the illuminated portion is then recorded by spatially resolved detection of a diffraction intensity of the illumination light diffracted from the illuminated portion in a detection plane. The steps of “illuminating” and “recording the diffraction image” are then carried out for further portions of the lithography mask. Between at least two portions of the lithography mask that are thereby detected, there is in each case an overlap region whose surface extent measures at least 5% or more of the smaller of the two portions of the lithography mask. The repetition takes place until the detected portions of the lithography mask completely cover a region of the lithography mask to be detected. The structure of the lithography mask is calculated from the recorded diffraction images of the illuminated portions. A device for carrying out the structure detection method is also specified, which comprises a light source, a spatially-resolving detector and a mask holder. This results in a method and a device for detecting the structure of a lithography mask, in which the demands placed on an optical unit arranged downstream in the beam path from the illumination light of the lithography mask are reduced.