BetaAbout 1 minute

AI Text Cleaner

Some AI-generated text may carry invisible Unicode characters or recognition markers. AI Text Cleaner helps you inspect supported findings and clean them before publishing.

Quick task summary

Best for:
Reviewing and cleaning hidden Unicode before publishing text
Input:
Pasted text or a text file up to 5 MB
Output:
A reviewable finding report and cleaned text
Next action:
Review preserved characters before publishing; coverage evolves as reliable text markers are verified
Quick Answers

TL;DR: Paste or import text, review hidden-character findings, and use Safe Clean by default. Compare the original and cleaned text before copying or downloading the result.

Loading tool
What the current Beta checks
Choose your Beta cleaning mode

Use Safe Clean for ordinary publishing workflows. Switch to Deep Unicode Clean only when you can review complex scripts, bidirectional text, emoji, and custom-font characters. The Beta focuses on recognition markers that can be located in the text itself.

Use this tool for

Invisible Unicode hygiene

Scan and clean supported hidden characters introduced by copy/paste, editors, documents, or Unicode steganography.

Default

Safe Clean

Remove high-confidence carriers while preserving contextual ZWJ/ZWNJ, RTL, mathematical, script, variation-selector, and flag-tag cases for review.

Use with care

Deep Unicode Clean

Apply stronger deterministic cleanup when you accept the risk of changing script-sensitive rendering.

Beta focus

Text-level hidden markers

The scanner works with supported invisible characters and recognition markers that can be located in the text itself. Coverage will be reviewed and updated as reliable signals are verified.

Separate workflow

Document and image files

This tool reads text files. Document and image inspection belongs in a separate file-review workflow.

Before publishing

Review the output

Compare the findings and cleaned text, then review meaning and formatting before you publish.

How to Use
1

Paste or import text

Paste text into Original Text or import a TXT, MD, CSV, or JSON file up to 5 MB.

2

Review the scan

Inspect counts, reveal readable hidden-character tokens, and open individual findings for code point, category, position, and recommendation.

3

Choose Safe or Deep cleaning

Use Safe Clean by default. Deep Unicode Clean is more aggressive and requires review of complex scripts, emoji, and bidirectional text.

4

Copy or download

Copy the read-only cleaned result or download it as a TXT file. The tool will evolve as reliable text-marker features and test results become available.

FAQ
What does this tool do?

Some AI engines may place invisible Unicode characters or recognition markers in generated text. This Beta scans supported hidden characters, explains the findings, and lets you clean them before publishing.

Which text markers does it check?

It checks supported hidden Unicode characters and recognition markers that can be found in the text itself, including zero-width characters, soft hyphens, bidi controls, unusual spaces, tags, and contextual script characters. Coverage may differ across AI engines, so review the result before publishing.

How will Claude-related text watermarks be handled?

Claude and other AI engines may add new text-watermark methods over time. We will keep evaluating reliable signals and update the supported rules as coverage is verified.

Why are some ZWJ or ZWNJ characters preserved?

ZWJ and ZWNJ can be required by family emoji and Arabic, Persian, Indic, Khmer, Mongolian, Hangul, or other complex-script text. Safe Clean preserves contextual cases for review instead of silently damaging them.

Will it change visible punctuation?

Only if you turn on optional typography normalization. Em dashes, curly quotes, and ellipses are ordinary visible typography and are left unchanged by default.

When should I use Deep Unicode Clean?

Only when you can review the result for complex scripts, bidirectional text, emoji, CJK or Ogham spacing, and custom-font private-use characters.

Can I import a document or image?

The Beta accepts TXT, MD, CSV, and JSON text files. Document and image inspection belongs in a separate file-review workflow.

Tool Overview
Expand to understand what this tool does and why it helps
Read Beta notes

Beta for hidden-character cleanup

Some AI-generated or copied text may contain invisible Unicode characters or recognition markers. AI Text Cleaner scans supported characters and applies context-aware Safe Clean rules for review before publishing.

What it helps with

  • - Reveal zero-width characters, soft hyphens, bidi controls, tags, unusual spaces, and private-use code points
  • - Protect emoji joiners and script-sensitive characters during Safe Clean
  • - Compare removed, normalized, and preserved findings before publishing
  • - Copy or download the cleaned result for your publishing workflow

Practical outcomes

  • - Cleaner product descriptions, ad copy, email copy, and other pasted ecommerce text
  • - Fewer hidden formatting surprises across editors and publishing systems
  • - A reviewable hidden-character report that can evolve as reliable text-marker patterns are tested

Useful tool? Share it with your team

Copy the link or share this tool page directly to your preferred channel.