Environmental assessments of regional development projects have been used in Mexico to determine where conflicts between conservation of biodiversity and resource extraction are likely to occur. Species‐rich areas have been acknowledged as a priority for conservation. However, biological information is incomplete and biased toward accessible sites, so species‐rich areas cannot be depicted directly from current biological knowledge. An alternative approach to predicting species‐rich areas is presented in this article. It is based on the gap analysis technique and involves the use of ordination analysis and generalized linear models integrated with a geographic information system. This approach was used for locating species‐rich areas in the Mexican states of Guerrero and Oaxaca, where a regional forestry development project was proposed. Baseline information consisted of geo‐referenced collection sites of terrestrial vertebrates. Thirty‐two species assemblages were identified by the ordination analysis, as well as by 25 generalized linear models. Validation of six of these models showed no significant differences between observed and predicted species frequencies. Results demonstrated that species‐rich areas could be depicted even under the constraints of environmental assessment in Mexico. A large number of species could be used in this analysis due to the minimal information required for each species record. This predictive approach optimized available biological information for the integration of conservation into regional development planning.