Your essay was flagged as AI and you wrote every word

Why clean human writing gets flagged

Detectors mostly measure two things. Perplexity is how surprising each word is given the words before it. Burstiness is how much your sentence length and complexity vary across a paragraph. Machine prose scores low on both — it is smooth and even.

So does a lot of good human writing. If you write in short, regular sentences, stick to plain vocabulary, follow a five-paragraph structure you were taught, or write carefully in a second language, you produce a low-variance text. The detector cannot tell that apart from a model, because on the axes it measures, they look the same.

• Writing in English as a second language — the best-documented false-positive group
• Formal, formulaic academic structure, especially taught templates
• Technical or legal subject matter where phrasing is conventional by necessity
• Heavy use of a grammar tool's rewrite suggestions rather than its corrections
• Short submissions — under roughly 300 words, every detector becomes unreliable

Before you reply to anyone, preserve your evidence

Do this first, today, before you write a single message. Evidence about how a document was written is fragile, and some of it expires.

• Do not edit, re-save or 'clean up' the document. Leave the file exactly as submitted.
• In Google Docs, open File → Version history → See version history. It holds a timestamped record of the document growing. This is the single strongest artefact most students have.
• In Microsoft Word, check File → Info → Version History if the file lived on OneDrive or SharePoint.
• Collect your notes, outlines, photos of handwritten planning, library loans, and any draft you emailed to yourself or a friend.
• Save messages where you discussed the assignment with anyone — a classmate, a tutor, the instructor.

Ask which detector produced the score, and what the score means

Different tools disagree with each other constantly. The same paragraph can score 8% on one detector and 91% on another. That disagreement is itself an argument, and it is easy to demonstrate.

Ask, politely and in writing: which tool was used, what the reported number was, and whether the institution has a documented threshold for acting on it. Many institutions have a policy that a detector score alone is not sufficient grounds for a misconduct finding — Turnitin itself says so in its own guidance to educators.

What actually persuades people

Not indignation, and not a technical lecture about perplexity. What works is a short, factual account plus artefacts they can open.

The strongest single move available to most students is to offer to discuss the content of the paper. If you wrote it, you can explain why you chose that source, what you cut and why, what you found hardest. Nobody who pasted a document can do that.

• A calm one-page timeline of how and when you wrote it
• A link or export of the version history
• Your outline and notes
• An offer to talk through the argument in person
• If relevant: a note that you are a non-native English speaker and that this is a documented false-positive factor

What to avoid

• Do not run your text through a humanizer and resubmit. If it is discovered, you convert a disputed score into a concealment problem — a far worse position.
• Do not rewrite the document to lower the score before the conversation. It looks like tampering and destroys the version history argument.
• Do not open with an accusation about the tool. Institutions defend their tooling; they respond much better to evidence about your process.
• Do not ignore it and hope. Deadlines for appeals are usually short and strictly enforced.