Effective conservation of (semi-)natural grasslands requires an understanding of the factors affecting naturalness (i.e. the actual quality of a habitat or vegetation patch) and the importance of the particular factors. Both local or patch and landscape or matrix variables affect habitat quality, and the proportions of the effects need to be identified. Therefore, we performed a hypothesis generating and testing analysis with generalised linear models on three typical grassland habitat types (forest steppe meadows,
alkali steppes, and lowland wet meadows), differing in their fragmentation, ecology and history, and representing characteristic types of grassland habitats with the use of the national database of the vegetation of Hungary (MÉTA). Our results, in general, show that naturalness depends upon both intra-habitat and matrix attributes: presence or proportion of other habitat types in the surrounding landscape, threatening factors and landscape ecological attributes. Higher number of habitat types and higher proportions of (semi-)natural habitats in the landscape have significant effects: presence of other grassland types similar in ecological demands to the model habitat positively affect the naturalness, while non-characteristic, secondary or disturbed habitats and invasive alien species have negative effects. However, there are clear differences among the three habitat types, indicating that for effective conservation, good knowledge of conserved habitat types is essential. Landscape or matrix factors, both compositional and structural, affecting habitat patch quality have significant effects that cannot be overlooked. In the case of fragmented grasslands, matrix factors might be even more important than patch or local factors.