The question of Caesarian authorship of De Bello Alexandrino, De Bello Africo, and De Bello Hispaniensi has puzzled classical scholars for millennia. Although these three texts have traditionally been attributed to Caesar, stylistic differences between these texts and the rest of Caesar’s corpus have resulted in doubts as to their true authorship. Prior computational studies on the authorship of these texts have involved traditional stylometric feature-based analysis and methods from the field of distributional semantics. However, with the advent of the transformer architecture in 2017, more sophisticated approaches to authorship verification have since emerged. This study builds upon previous results by employing the state-of-the-art transformer-based model Siamese BERT to investigate the authorship of De Bello Alexandrino, De Bello Africo, and De Bello Hispaniensi. As a secondary goal, this study also evaluates the effectiveness of Siamese BERT in conducting authorship analysis in Latin, thus assessing its potential for language-agnostic application. Following training on an open-source dataset provided by Vainio et al. (2019), the model achieved a 95.5% accuracy on the validation dataset. The model was further validated through cross-comparison of each text in Caesar’s De Bello Gallico corpus, successfully identifying Book VIII as an outlier with regards to authorship. Finally, following authorship verification of the unknown texts using the model, results suggest that De Bello Alexandrino and De Bello Africo may have been written by Caesar, whereas De Bello Hispaniensi definitively was not written by Caesar. These findings offer new insight into a two-millennium-old mystery of authorship and contribute meaningfully to the historical interpretations about these texts. Furthermore, the results demonstrate that Siamese BERT is effective in authorship analysis of Latin, indicating its broader applicability across languages.