This is the site-level data used to analyses whether sampled biodiversity differs between sites inside and outside protected areas. These data were obtained from the November 2014 version of the full Projecting Responses of Ecological Diversity In Changing Terrestrial Systems (PREDICTS) database (described in Hudson et al. 2014 The PREDICTS database: a global database of how local terrestrial biodiversity responds to human impacts Ecol. Evol. 4: 4701-4735 DOI: 10.1002/ece3.1303), but include only studies that had sites both inside and outside of protected areas (as described by IUCN and UNEP (2014) The World Database on Protected Areas, July 2014. UNEP-WCMC, Cambridge. Available at www.protectedplanet.net). The data contain the source, study, block and site IDs. For each site the following variables are given: country, latitude, longitude, diversity measure (abundance or occurrence), land use (both as an eight level and three level factor), use intensity, the start date of sampling, the realm, latitudinal zone, broad taxonomic group, the four response variables analysed (abundance, species richness, rarefied species richness and community weighted mean range size), data on covariates (slope, elevation and agricultural suitability) and whether each site fell within a protected area or not. For each protected site, the following variables from the July 2014 World Database on Protected Areas are provided: IUCN category (1.5 = I & II, 4.5 = III to VI, 7 = unreported), the GIS_AREA, STATUS_YEAR, duration of protection (status year minus earliest sampling date at that site), and the size/age class of the protected area. Variables are also included to specify which sites were used in the matched landuse analyses and the species richness/rarefied richness models (which included only sites for which multiple taxa were surveyed).