In the genomics era, bioinformatic analysis, especially in non-model species, facilitates the identification and naming of numerous new proteins, the function of which is then inferred through homology searches. Here, we question certain aspects of these approaches. What are the criteria that permit such a determination? What are their limits? Naming is classifying. We review the different criteria that are used to name a protein and discuss their constraints. We observe that the name given to a protein often introduces a bias for further functional analyses, a bias that is not often taken into account when analysing results. Last but not least, the heterogeneity of criteria used for naming proteins leads to self-inconsistent or contradictory protein classification that is potentially misleading. Finally, we recommend a wider use of phylogenetic criteria in protein naming.