The theme in this dissertation is the interpretation and reproducibility of results in fluorescence microscopy experiments using fluorescent protein tags. Scientists often assume that fluorescent protein tags are inert, however, as we show, these tags can cause problematic experimental artifacts. In the first chapter, we fused fluorescent protein tags to reovirus µNS to study the dynamic properties of fluorescent viral factory-like structures. We found that the tags promoted biomolecular condensation of reovirus µNS and that the tags also influenced properties like the number and size of µNS condensates. After screening nine tags, we identified mNeonGreen as a suitable tag for the N-terminus of µNS, because it least affected the condensation properties that we studied. Another tool used in fluorescence microscopy is the representative image, which is a visualization tool used to communicate research results in static figures. In the second chapter, we demonstrated an objective method of image selection that used an autoencoder model, which we compared to a published method, but we found that both methods were limited by their reliance on labels. In the third chapter, we turned to an alternative visualization tool, virtual microscopy, to stream interactive datasets to reader’s browsers on the web. We applied virtual microscopy to replicate highly magnified representative images from an original experiment that used mCherry fusion proteins to visualize the endoplasmic reticulum during reovirus infection. I hypothesized that the fluorescent tag structured the organelle, not reovirus.