Draft Notes
Short practical notes on editing AI-assisted drafts, reading AI detector reports and keeping your own voice in academic and professional writing.
- Why AI Detectors Keep Flagging Your Conclusion — Conclusions are the most uniform paragraphs in most essays, which is exactly what a detector notices. Here is the edit that fixes it.
- The 'Sounding Like a College Student' Trap — Aiming for a generic student voice produces the most machine-like prose. Write like one specific person instead.
- Paraphrasing a Definition Does Not Remove the Flag — A definition swapped word for word keeps the same structure and the same risk. Rewrite the shape, not the vocabulary.
- Editing a Literature Review for Rhythm, Not Vocabulary — Literature reviews are the hardest section to make sound human. The fix is sentence variety and real transitions, not bigger words.
- What Changes When You Cut 'However' — Transition words are a crutch. Removing them forces the sentence to show the relationship instead of naming it.
- The Three-Paragraph Test Before You Submit — Read the first, longest and last paragraphs of your essay. They carry most of the detector signal and most of the reader's judgement.
- Reading a Turnitin AI Report as a Student — The score is a segment estimate, not a verdict. Read the flagged segments, find the uniform sentences, fix those.
- What to Say When a Teacher Asks About AI Use — A calm, honest script for the conversation after a flagged essay, plus the evidence worth keeping before it happens.
- Detector Scores and Group Work: Whose Voice Is It? — Group essays are flagged more often because the draft has no single voice. One editing pass fixes most of it.
- Writing a Personal Statement Without the Robot Cadence — Personal statements fail detectors differently: they are too smooth. The cure is one specific scene, not more adjectives.
- Why the Abstract Trips Detectors More Than the Body — Abstracts compress and nominalise by design. A few concrete anchors stop a detector (and a reader) from finding them generic.
- Three Sample Sentences That Read Human, and Why — Concrete examples of the cadence edits that make prose read like a person wrote it, with the reasoning spelled out.
- Formatting Is Not Voice: Why Headings Do Not Fix AI Prose — Bold labels and tidy headings make a document look organised while the sentences stay uniform. Detectors read the sentences.
- Two Tools, Two Jobs: Humanizer vs Detector — One tool rewrites, the other estimates. Using them in the wrong order wastes a revision cycle.
- The Fastest Fix for a Uniform Paragraph — A five-minute method: split one sentence, merge two, add one concrete detail, read aloud. That is usually enough.
- What 'Acceptable AI Percentage' Actually Means on a Rubric — Policies that quote a number are quoting an estimate. Read the wording, keep the evidence, and treat the number as a signal.
Tools mentioned in these notes: Better Humanizer, a free AI humanizer with a built-in detector, and its AI detector.