So far, few studies have been able to account for the issue of the artistic phenomenology and fruition. A semiotic analysis of current computer-generated images offers the extraordinary chance to learn more about personal and public reactions to masterpieces of art. In fact, just by entering in Google the keywords “Mona Lisa”, “Starry Night”, “Venus of Milo”, or by running the same prompts in Midjourney or Stable Diffusion, it is possible to bring to the surface and surprisingly discover one individual and collective elaboration of these and other similar famous masterworks. Both the large collections of images stored in databases and images that are produced through AI tools are a rich and clear testimony of people’s different proxemic relations with great works of art and of scopic regimes of enunciation, from very intimate close-ups to social and political wide-angles. We will consider the results of such algorithmic processes as intersemiotic translations between the natural language and the visual one. Above all, by applying Jacques Fontanille’s model of six levels of immanence – signs, texts, objects, practices, strategies, and forms of life – we will try to test the veridictive status of these generated images for the understanding of the dynamics of cultural inheritance.