Persons and affiliations

Date

Description

Post-editing of texts written with generative AI tools (e.g., the GPT suite, Gemini, Claude, and QuillBot) is a task that can be performed by humans, machines, or a combination of both. Currently, artificial post-editing is gaining momentum with diverse actors (from students to journalists), employing it to achieve different goals. These tools, known as text humanizers, claim to add a “human touch” to AI-generated output by eliminating its robotic “tone” and making the result more natural, human-like, and authentic. Building on previous corpus research (e.g., De Cesare 2023 on biographies), this talk aims to: (i) highlight key linguistic and textual features characteristic of LLM-generated outputs; (ii) examine whether and how these features are modified by text humanizers; and (iii) describe the typical changes that occur in artificially post-edited texts. Are these changes primarily lexical, involving substitutions of common words with more specialized vocabulary? Or do they extend to syntactic and textual structures as well?