Policies aimed at sustainable landscape management recognise the importance of multiple cultural viewpoints, but the notion of landscape itself is implicitly assumed to be homogeneous across speech communities. We tested this assumption by collecting data about the concept of "landscape" from speakers of seven languages of European origin. Speakers were asked to freely list exemplars to "landscape" (a concrete concept for which the underlying conceptual structure is unclear), "animals" (a concrete and discrete concept) and "body parts" (a concrete concept characterised by segmentation). We found, across languages, participants considered listing landscape terms the hardest task, listed fewest exemplars, had the least number of shared exemplars, and had fewer common co-occurrence pairs (i.e., pairs of exemplars listed adjacently). We also found important differences between languages in the types of exemplars that were cognitively salient and, most importantly, in how the exemplars are connected to each other in semantic networks. Overall, this shows that "landscape" is more weakly structured than other domains, with high variability both within and between languages. This diversity suggests that for sustainable landscape policies to be effective, they need to be better tailored to local conceptualisations.