This detector is designed to be readable and auditable. You get an overall interpretation, but also the sentence-level evidence and reason keywords that influenced the result.
Drop in an essay, article, or paragraph. Live stats keep word and sentence counts visible before you run analysis.
The detector checks vocabulary, structure, sentence rhythm, and common AI-like phrasing patterns across the text.
Each sentence gets a signal level so you can see where risk is concentrated instead of relying on one overall number.
Use the report summary, signal cards, and evidence details to support a careful, human judgement.
Every scan classifies each sentence as human-like, mixed, or AI-like, highlights it in your document, and names the pattern behind every flag with reason keywords. You can export the whole report as PDF or DOCX.
Run an analysis on the Detector page, then use the "View full report" button in the result banner. The report stays available until you clear or replace the analysis.
We don't reduce detection to a single unverifiable percentage. Every scan classifies each sentence and names the pattern behind any flag, so you can check the result against the text yourself instead of trusting a bare score.
The detector evaluates writing patterns and structural signals common across modern AI-assisted drafting styles.
Yes — as evidence inside your process, which is exactly what the sentence-level view is built for: a report you can point to, with the flagged sentences and the reason named for each. It is never the whole case on its own. Pair it with revision history and a conversation with the student, the way any evidence in a fair process gets weighed.
This version does not keep a remote database of your text; current-session and local browser state are used for usability.
Run analysis, inspect flagged sentences, review reason keywords, and combine results with author context.
Paste any text and get the same sentence-level breakdown you just read about. Free, no signup required.