Winston AI checks French writing for signs of AI, from a lycée dissertation to a Montreal agency's blog post, and marks the sentences that read as machine-written. Here is what French writing conventions mean for a result, where tu and vous fit in, and how to keep a score in perspective.
The dissertation problem
French students are trained in a very particular form. The dissertation follows a plan (often thèse, antithèse, synthèse), opens with an accroche and a problématique, and moves between parts with explicit transitions. The commentaire de texte has its own fixed structure too. That training produces essays that are tidy, balanced and full of connectors, which also happens to describe how default AI output tends to read.
So structure alone tells you very little. A well-taught lycéen and a chatbot can both write «Dans un premier temps, nous verrons...». Look instead at what the highlighted sentences say. Analysis that never quotes the text, or a conclusion that could close any essay on any subject, gives you something concrete to discuss. A neat plan does not.
- Are the examples and quotations specific to the text being studied?
- Does the problématique match the plan the student sketched in class?
- How do the flagged parts compare with the student's devoirs sur table, written by hand in class?
Formulas outside the classroom
Professional French has its own scripts: the cover letter that opens with «Je me permets de vous adresser ma candidature», the report written in a careful nominal style, the administrative email that closes with a full formule de politesse. These are human conventions, and plenty of honest writing leans on them. Text from France, Quebec, Belgium or Switzerland is all French to Winston AI and is detected automatically, so paste it as-is.