Text watermark checker for hidden Unicode and formatting artifacts

A text watermark checker can inspect characters that normal rendering hides. Scan AI Text reports deterministic evidence and keeps unsupported claims out of the result.

What the checker looks for

• Zero-width spaces, joiners, and word joiners outside supported context.
• No-break, thin, figure, ideographic, and other unusual space characters.
• Bidirectional marks, isolates, embeddings, and overrides.
• Unicode tag characters and detached variation selectors.
• Private-use, noncharacter, reserved-ignorable, and other format codepoints.
• Optional lookalike Latin characters when an aggressive review is explicitly requested.

Why context-aware cleaning matters

Blindly deleting every invisible character can break emoji, multilingual spelling, flags, glyph variants, and right-to-left text. Scan AI Text uses conservative context rules adapted from the MIT-licensed watermarks-remover project. It reports more than it removes, so a review can distinguish an artifact from meaningful text structure.

Is this an AI text watermark detector?

It is an AI text watermark checker only for directly inspectable text artifacts. It is not a universal detector for undisclosed token-sampling watermarks, vendor systems, or AI authorship. Any writing-pattern hints appear separately and carry a low-confidence label.

What happens to uploaded text?

Supported files are read by the browser. Artifact scanning and cleaning happen on the device. Optional analytics events never include pasted text, filenames, cleaned output, codepoints, or document contents.

[Open the scanner](/) or read about AI text watermarks.