Claude Text Cleaner

Invisible combining mark guide · October 30, 2026

Does Claude Text Cleaner Remove Combining Grapheme Joiner?

No. This browser-only cleaner does not remove U+034F COMBINING GRAPHEME JOINER (CGJ) from pasted text. In a live cleaner test, e + U+034F + U+0308 stayed three code points and the report showed zero removals. A zero count means none of this cleaner's supported categories matched; it does not mean every invisible character is absent.

Try the browser cleaner

The labels show the logical code-point sequence. The mark itself is invisible, so appearance alone cannot confirm whether it is present.

Why might CGJ be there?

The Unicode Standard, chapter 23 §23.2.4, describes CGJ as a way to affect language-sensitive collation and distinguish sequences that would otherwise be canonically equivalent. It can matter to normalization and searching even though it has no visible glyph. A public Stack Overflow question asks why a decomposed e plus diaeresis remains after Unicode normalization when U+034F appears between them. That question is about Unicode processing, not this site's cleaner.

This cleaner has fixed removal categories: selected zero-width characters, bidirectional controls, tag characters, variation selectors, private-use characters, soft hyphens and em dashes. U+034F is in none of those rules. The cleaner neither normalizes text nor identifies why a particular mark was inserted.

What to do when a zero count leaves a text mismatch

  1. Keep the original. Make a copy before cleaning, especially for names or multilingual text.
  2. Check the cleaner report. Paste the copy into the browser cleaner. Zero removals means its supported rules made no deletion.
  3. Inspect code points separately. If search, sorting or exact-string matching still behaves unexpectedly, use a Unicode code-point inspector or a code-point dump in your editor to locate U+034F. Decide whether it is intentional for that language or workflow before changing the source text.

The three-code-point result above was tested in the live browser cleaner on October 29, 2026. It does not establish how another application normalizes or renders the sequence. For another retained invisible character, see the U+2063 invisible separator guide. The cleaner does not establish text origin, detect a watermark or predict an AI detector result.