Gemini lives inside tools people already use every day, like Gmail and Google Docs, so its text often arrives without anyone opening a chatbot. Winston AI flags machine-written text, including text from Gemini, and marks the sentences that read as AI. Here is what that looks like in shared documents and how to check them fairly.
AI help built into the document
When the writing assistant sits inside the editor, the line between drafting and editing gets blurry. A person might ask Gemini to expand three bullet points into a paragraph, make a sentence sound more formal, or draft a reply to an email thread, all without leaving the page. The result is a document where AI text and human text sit side by side, sometimes in the same paragraph.
That pattern is easy to see in sentence highlights. Expect pockets of red where a section was generated, yellow where a person's sentence was reworked for tone, and plain text where they typed it themselves.
Use the document's own history
Google Docs keeps a version history, which is some of the best context you can get. If a scan raises questions, look at how the document grew: steady typing over several sessions looks different from large blocks appearing at once. Pasting can be innocent too (notes from another file, a quote, a template), so treat history as context, not a gotcha.
Gemini in school accounts
If your school uses Google Workspace, students may have Gemini close at hand, depending on how it has been set up. Agree on which uses are fine (brainstorming, grammar fixes) and which are not, then check work against that policy. Our page for teachers and the guide to how AI detectors work go into more detail.