This research compares expert and non-expert conceptualisations of land cover data collected through a Google Earth web-based interface. In so doing it seeks to determine the impacts of varying landscape conceptualisations held by different groups of VGI contributors on decisions that may be made using crowdsourced data, in this case to select the best global land cover dataset in each location. Whilst much other work has considered the quality of VGI, as yet little research has considered the impact of varying semantics and conceptualisations on the use of VGI in formal scientific analyses. This study found that conceptualisation of cropland varies between experts and non-experts. A number of areas for further research are outlined.