Abstract
This paper analyzes the discrepancies between the historical reality of certain poems written decades or even centuries ago, and the results provided by modern programs designed to detect AI (Artificial Intelligence) generated content. These discrepancies are important because they can lead to errors in the authentic evaluation of literary texts affecting both the recognition of the cultural value of these works and the editorial process. Such a phenomenon can contribute to a distorted perception of classical authors and a lack of confidence in the validity of literature transmitted over the centuries.
In the academic environment, some classical texts analyzed using such AI detectors have been erroneously marked as artificially generated, which has led to unnecessary re-evaluations of their authenticity or even the rejection of academic work. Through a case study on several classical and contemporary poems scanned with such a tool, we observe how AI detectors can incorrectly classify authentic texts as AI-generated. These errors raise issues concerning the use of such instruments in evaluating the originality of literary content especially in academic and editorial contexts.