Grizzly bear (Ursus arctos) habitat use maps delineate predicted habitat use for grizzly bears around federally designated recovery zones in and near western Montana. These raster data are the official data release for Sells et al. (2022), “Grizzly Bear Habitat Selection Across the Northern Continental Divide Ecosystem” and Sells et al. (2023), “Grizzly bear movement models predict habitat use for nearby populations.” In summary, to better understand habitat selection by grizzly bears, Sells et al. (2022) developed and validated individual-based integrated step-selection functions (iSSFs) for 65 grizzly bears monitored in the Northern Continental Divide Ecosystem (NCDE; including 19 males and 46 females). Step-selection functions, developed from GPS location data, are highly versatile models that can be used to study and predict habitat selection, movement corridors, dispersal, and human-wildlife interactions. Sells et al. (2022)’s models involved variables that could be measured across a large landscape and were previously identified as important to habitat selection by this generalist omnivore. External validation with data omitted from model development demonstrated that the models were highly predictive of habitat use within the NCDE. Sells et al. (2023) built on Sells et al. (2022) as a second phase of the study. In this paper, we aimed to evaluate the transferability of the models to a larger landscape spanning multiple ecoregions. We simulated movements within the Selkirk (SE), Cabinet-Yaak (CYE), and Greater Yellowstone (GYE) Ecosystems using the 65 models Sells et al. (2022) developed from NCDE bears, summarized results to estimate relative habitat selection, and assessed model transferability over space and time using location data for bears in the SE, CYE, and GYE. Because the NCDE models can be accurately transferred to nearby populations, this demonstrates reliability of applying the models to predict habitat use for areas with few or no data, including potential connectivity pathways between populations or in uninhabited areas like the North Cascade (NCE) or Bitterroot Ecosystems (BE). In these datasets, all results are summarized and quantile binned as classes 1 (lowest relative predicted use) - 10 (highest relative predicted use) for each recovery zone. Each class is represented by 10% of each recovery zone. However, because the NCDE-CYE and SE-CYE slightly overlap, we averaged and rounded to the nearest integer at cells of overlap. The maps can be clipped down as desired to view details of specific areas. (Note: subsetting the maps will mean that the classes no longer each represent 10% of the landscape, as this binning was done independently for the NCDE, SE, CYE, and GYE.) Our predictive maps can facilitate on-the-ground application of this research for prioritizing habitat conservation, human-bear conflict mitigation, and transportation planning.