Palaeontological field surveys in remote regions are a challenge, because of uncertainty in finding new specimens, high transportation costs, risks for the crew and a long time commitment. The effort can be facilitated by using high-resolution satellite imagery. Here we present a new opportunity to investigate remote fossil localities in detail, mapping the optical signature of individual fossils. We explain a practical workflow for detecting fossils using remote-sensing platforms and cluster algorithms. We tested the method within the Petrified Forest National Park, where fossil logs are sparse in a large area with mixed lithologies. We ran both unsupervised and supervised classifications, obtaining the best estimations for the presence of fossil logs using the likelihood and spectral angle mapper algorithms. We recognized general constraints and described logical and physical pros and cons of each estimated map. We also explained how the outcomes should be critically evaluated with consistent accuracy tests. Instead of searching for fossiliferous outcrops, our method targets single fossil specimens (or highly condensed accumulations), obtaining a significant increase in potential efficiency and effectiveness of field surveys. When repeatedly applied to the same region over time, it could also be useful for monitoring palaeontological heritage localities. Most importantly, the method here described is feasible, easily applicable to both fossil logs and bones, and represents a step towards standard best practices for applying remote sensing in the palaeontological field.