Images from Ancient Greek: Towards a Semantic Reflection on ChatGPT’s Text-to-Image Generation

The creation of images using ChatGPT, Gemini, and other chatbots currently represents one of the most widespread applications of Generative Artificial Intelligence. This article originates from a personal experiment: ChatGPT (4o model) was asked to produce an image based on a passage from the “True History” by Lucian of Samosata, first using an Italian translation and then the original Greek text. The comparison between the results raised questions concerning the model’s ability to “read” Ancient Greek and to associate certain terms with images corresponding to specific real-world objects. Several months later, the experiment was repeated using a more up-to-date version of the model (GPT-5.1), in order to monitor changes, possible improvements and variations. Building on this case study, the article proposes a reflection on the pedagogical potential of chatbots in the study of the Greek language – particularly regarding vocabulary and its semantic nuances – and, more broadly, on the complexity of the translation process from Greek into Italian.

Full-text of the article is available for this locale: Italiano.