How to show you wrote your essay

A detection score describes how text reads, not where it came from, which cuts both ways. What carries weight is the record of the work, and most of it has to exist before you need it.

The awkward property of an AI score is that it says nothing about process. It reads the finished words. That is what makes a flag frustrating, and it is also the opening: everything about how the document came to exist sits outside what the classifier can see, and it is all yours.

Most of it has to be collected before you need it, which is the reason to read this now rather than later.

What exists already, if you know where to look

Version history. Google Docs keeps revision history indefinitely. Word tracks versions through OneDrive and SharePoint. This is the strongest single record there is, because it shows the document growing over hours rather than appearing whole.

Drafts in your email. Anything you sent to yourself, a supervisor, a study partner, with timestamps attached.

Comments and feedback. Supervisor notes on an earlier draft, tutor comments, peer feedback. Each one is dated evidence of a document that existed in an earlier state.

Notes and reading. Annotated PDFs, a reference manager library, handwritten notes photographed on your phone, the browser history of the sources you used.

Your own earlier work. Prose you wrote before any of this was a question establishes what your writing normally looks like, which is useful when the claim is that this piece does not sound like you.

Build the habit now, not after

Write in something with version history. Google Docs or Word with autosave to OneDrive. This costs nothing and produces the record automatically.

Do not compose in a plain text editor and paste in at the end. A document that appears in one paste has no history to show.

Keep the notes. The intermediate mess is worth more than the polished file.

Understand what the flag was reacting to

This is the part most people skip, and it is the one that turns a defensive conversation into a specific one.

A score is a reaction to measurable properties of your prose. Run the document through a second detector, per paragraph, and you can see which passages a classifier reads as flat and why. "This section is a methods description written to a convention, and here is what that does to a classifier" is a concrete observation. "I did not use AI" is a position.

The properties are known and countable. Sentence length variation: human academic writing in our corpus has a standard deviation of 8.7 words, generated text 6.4. Connective openings: 2.4% of human sentences against 5.8% of generated ones. If your writing sits near the generated end on these, you can show that, and you can show it is a property of the register rather than the origin. See why AI detectors flag human writing.

Numbers worth having to hand

Turnitin's published wrong-flag rate on human writing: 4%
Turnitin's own finding that documents scoring below 20% show a higher incidence of wrong flags
Turnitin's own finding that 54% of falsely flagged sentences sit directly beside a

genuinely flagged one, and 26% two sentences away

Elevated wrong-flag rates for non-native English writers, found repeatedly in published

evaluations. See why non-native English writing gets flagged

What follows from a score is governed by your institution's own procedures, which vary and which we are not the right people to describe.

The cheapest version of all of this

Check before you submit. A document you have already looked at through a detector is one where you know which paragraphs are exposed and have had the chance to fix them.