The development of a species-specific conservation protocol that involves artificial insemination with frozen semen needs to validate an effective methodology for freezing semen. Colloid centrifugation has been suggested and widely applied as an effective tool for selecting animal spermatozoa for artificial breeding. The objective of the present study was to compare different methods of centrifugation, single layer using Androcoll-Bear and Percoll and double layer using PureSperm 100 (in two different discontinuous gradients 40%-80% and 45%-90%), for the selection of fresh brown bear sperm samples. In the before freezing group, all selected samples showed a higher progressive motility and viability (except Percoll for motility 43.0 ± 5.3 [P < 0.05]); all colloids except PureSperm 45/90% rendered samples with fewer damaged acrosomes. In the after thawing group, all tested centrifugation colloids showed a good capacity to decrease the number of damaged acrosomes. Furthermore, PureSperm treatment (45/90%) resulted in an increase in apoptotic-like changes not only immediately after thawing but also after the incubation test, leading us to suggest that this gradient could induce some kind of deleterious effects on the sperm samples. On the other hand, PureSperm treatment (40/80%) yielded a quality preservation capacity similar to Androcoll-Bear in number of damaged acrosomes, different relative to the control (control, 5.3 ± 0.6; PureSperm 80, 2.0 ± 0.3; Androcoll, 2.1 ± 0.9 [P < 0.05]) but a decrease in the number of viable spermatozoa recovered after thawing relative to the control (control, 21.2 ± 3.1; PureSperm 80, 13.7 ± 2.7 [P < 0.05]). In conclusion, Androcoll-Bear constitutes a useful tool for handling of brown bear ejaculates owing to its simple handling and procedure with a reliable sperm selection and freezability. This colloid yielded an improvement in several sperm parameters in brown bear frozen-thawed semen; the selected spermatozoa of fresh samples with this colloid showed a better resistance to freezing compared with the control sample not only for motility but also for viability.