How Scan AI Text works

Scan AI Text is a transparent text watermark checker for artifacts you can inspect directly. It reads pasted or uploaded text in your browser, identifies unusual Unicode and formatting controls, and explains every finding before it changes anything.

1. Add text

Paste text or upload a plain-text, Markdown, CSV, or HTML file. The artifact scan runs locally in your browser. Scan AI Text does not need to send the document to a model or remote detector.

2. Review evidence

Each result includes the exact Unicode codepoint, character offset, count, proposed action, and reason. Common findings include zero-width characters, unusual spaces, detached variation selectors, directional overrides, Unicode tag characters, and private-use codepoints.

Context matters. Some invisible characters are legitimate parts of emoji, Persian and Indic orthography, CJK or Mongolian variation sequences, flags, bidirectional prose, shorthand, music, or hieroglyph layout. Conservative cleaning preserves those cases by default.

3. Clean supported artifacts

The cleaner removes detached invisible carriers and normalizes unusual spaces. A before-and-after preview remains available before you copy or download the cleaned text. Preserved items stay in the output and remain listed for review.

What this workflow cannot prove

This workflow does not provide universal detection for secret statistical or model-level watermarks. Those systems can require the original generation key, tokenizer, algorithm, or a vendor detector. It also cannot determine AI authorship from prose alone.

Writing-pattern hints are separated from deterministic evidence and labeled low confidence. Treat them as editing observations, never as academic-integrity proof.

Read product limitations or [open the text watermark checker](/).