Preparing your local scanner…
The scanner runs in this browser, so your passage stays on this device.
Preparing scan GitHub
Review writing with evidence

See the patterns
behind the writing.

Paste a passage and get a transparent estimate of how its writing patterns compare with AI-generated text—with sentence-level evidence to guide your review.

Local analysis · Evidence you can inspect

Start with a sample
Add text to review Ctrl+Enter to scan
Drop to import

For a steadier estimate, use 3+ sentences or 150+ words.

Ready for your passage.

0 / 8,000
0 words
Your review will appear here

Scan a passage to see an overall estimate, sentence-level evidence, and the writing patterns behind it.

About this scan

How it works, what the estimate means, and where it can fall short.

Where is my passage processed?

The scanner runs in this browser. Your passage is not sent to a detection API, and recent scans stay only in this tab’s session storage.

How should I interpret the estimate?

Treat the percentage as a signal to investigate, not proof of authorship. The 99.4% reference metric comes from a labeled held-out split; performance can change with passage length, topic, editing, translation, and paraphrasing.

What should I scan?

Use at least 3 full sentences; 150+ words usually gives a steadier estimate. Very short passages do not contain enough variation for the statistical signals to separate cleanly.

What do the sentence highlights show?

Each highlight shows how strongly a sentence aligns with the model’s AI-like or human-like patterns. Use the highlights to understand the estimate, not to make a high-stakes decision from one sentence.