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.
Example walkthrough
Overall score: 87.3% AI-Generated
88% - Likely AI
Artificial intelligence can rapidly improve productivity by automating repetitive administrative work.
54% - Mixed
Still, the outcome depends on clear instructions, human review, and context-specific edits.
19% - Likely Human
I rewrote this section twice after feedback because the first version sounded too generic.
Step 1
Drop in an essay, article, or paragraph. Live stats keep word and sentence counts visible before you run analysis.
Step 2
The detector checks vocabulary, structure, sentence rhythm, and common AI-like phrasing patterns across the text.
Step 3
Each sentence gets a signal level so you can see where risk is concentrated instead of relying on one overall number.
Step 4
Use the report summary, signal cards, and evidence details to support a careful, human judgement.
Use the final verdict only after reading sentence-level evidence and flagged signal categories.
For formal decisions, combine this output with document history, author context, and manual review.
Start with evidence, not an accusation.
See which sentences were flagged and why, then use the report alongside drafts, revision history and a conversation with the student.
Check your writing before you submit.
See which parts of your draft may read as AI-like and why, and make your writing clearer and more specific.
Scan papers without preparing them first.
Long academic documents are analysed in one pass, with references, citations and publishing boilerplate set aside automatically.
Screen submissions with evidence you can share.
Move from a single percentage to specific passages and reasons, in a report you can attach to the editorial record.
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.
No detector is, and any tool claiming 99.9% deserves scepticism. We show our reasoning sentence by sentence, so you can check us.
Text written with ChatGPT, Claude, Gemini and similar assistants. ScribeLens detects the patterns language models share, so a result tells you whether a passage shows AI-like patterns, not which tool produced it.
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.
English. The detection models and the pattern checks behind every sentence flag are built and calibrated on English writing, so results on other languages are not reliable — we would rather say so than hand you a confident-looking score. Text in any language is accepted, but only English text gets a meaningful report today.
Your analysis is kept in your own browser session so a refresh does not lose it, and it is cleared when the tab closes. A copy is only stored on our side if you explicitly create a verification link, and those expire automatically after 180 days.
Yes. The ScribeLens AI Detector extension for Chrome checks text without leaving the page. Select text on any web page and right-click "Check with ScribeLens": a side panel shows the overall result, the share of human, mixed and AI-like sentences, and the sentences most worth a look, with the reason for each. Up to 2,000 words are checked in the panel, with the same limits and privacy as the website. Longer text opens on scribelensai.com, where a free account checks the whole document, and "View full report" opens the complete report there without checking the text again.
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.