The United States Fish and Wildlife Service uses counts of unduplicated female grizzly bears (Ursus arctos) with cubs-of-the-year to establish limits of sustainable mortality in the Greater Yellowstone Ecosystem, USA. Sightings are clustered into observations of unique bears based on an empirically derived rule set. The method has never been tested or verified. To evaluate the rule set, we used data from radiocollared females obtained during 1975–2004 to simulate populations under varying densities, distributions, and sighting frequencies. We tested individual rules and rule-set performance, using custom software to apply the rule-set and cluster sightings. Results indicated most rules were violated to some degree, and rule-based clustering consistently underestimated the minimum number of females and total population size derived from a nonparametric estimator (Chao2). We conclude that the current rule set returns conservative estimates, but with minor improvements, counts of unduplicated females-with-cubs can serve as a reasonable index of population size useful for establishing annual mortality limits. For the Yellowstone population, the index is more practical and cost-effective than capture-mark-recapture using either DNA hair snagging or aerial surveys with radiomarked bears. The method has useful application in other ecosystems, but we recommend rules used to distinguish unique females be adapted to local conditions and tested.