<p>Online citizen science platforms provide georeferenced observations that are spatially extensive and may offer important insights into variation in individual traits. However, their utility for estimating phenotypic variation is largely untested. I examined iNaturalist data to estimate the frequency of coat colour polymorphism in American black bears (<i>Ursus americanus</i>) using two pairs of independent datasets. First, I used a roadside survey along a 493-km transect in the northern Rocky Mountains and compared results from iNaturalist along the same transect. Second, I compared data from 2,279 black bear mortalities in the Yukon with that from iNaturalist photographs from the same area. Of the two datasets in the northern Rocky Mountains the odds of observing a non-black colourmorph was greater in the iNaturalist dataset (odds ratio = 3.26); however, this difference was not statistically significant. Similarly, the odds of observing a non-black colourmorph in the two Yukon datasets was also greater in the iNaturalist dataset (odds ratio = 1.46), which was significantly different. The absolute percent difference in non-black colourmorphs was 8% in the northern Rocky Mountains dataset (4% in a systematic survey vs. 12% in iNaturalist) and 9% in the Yukon dataset (36% vs. 45%). The discrepancy between systematic and citizen science data in these two comparisons may reflect observer preference for black bears with atypical colouration. Alternatively, it may be due to initial confusion with sympatric grizzly bears (<i>Ursus arctos</i>), which are rarer and likely more desirable to photograph. These results provide new insight regarding the use of citizen science data to examine colour polymorphism. My findings indicate that caution is warranted when using citizen science data to estimate the frequency of phenotypic variation without independent validation. Similar studies to assess the generality of these results would be of value.</p>