Diagnostic word highlighter

When you are reviewing someone else's work, or your own draft after a day away, you do not want to start editing. You want to see. A read-only view keeps the text stable while you read it, so the thing you are judging cannot change under your cursor, and every mark means the same thing it meant a second ago.

It is also the honest way to review a model's output. Marking the text without touching it makes the density visible before any rewriting starts, which is the number that predicts how much editing the piece will actually need.

One flagged word is noise. Twenty in eight hundred words is a signature. The panel under the preview reports flags per 1,000 words so you can compare two drafts of different lengths on the same scale, and so a long document does not look worse simply because it is long.

Under three flags per thousand reads as ordinary professional prose; plenty of careful writers use these words. Between three and eight, readers start to feel the pattern even if they cannot name it. Above eight, swapping individual words will not help — the paragraphs need rebuilding, because the vocabulary is a symptom rather than the cause.

None of these lists prove a machine wrote the text, and this tool will never say it did. Human writers use every word here, and detectors that claim otherwise fail on careful writers, non-native speakers and anyone with a formal style. The list is an editing instrument: it points at the places where a sentence could be more specific.