Report reading guide · October 19, 2026

How to Read Claude Text Cleaner Removal Counts

“Removed characters” counts each supported Unicode code point deleted. “Changed categories” counts the categories with at least one deletion. Three removals can therefore belong to two categories. Original and clean lengths count code points before and after cleaning, even when the text looks nearly identical. These numbers describe this browser-only edit; they do not identify a watermark or the text's author.

Try the browser cleaner

The labels stand in for invisible code points so you can inspect the example. The actual input contains the characters, not the written labels.

Reproduce the mixed-character example

Select and copy this code span exactly into the tool: A​​B­C. Do not retype the visible letters alone. You can also construct the input as "A\u200B\u200BB\u00ADC" in a Unicode-aware editor. The visible letters are A, B, and C. This cleaner produces ABC: original length 6, clean length 3, removed characters 3, changed categories 2. Its category cards show “Zero-width characters” 2 and “Soft hyphen” 1. Keep the original if any hidden formatting was intentional.

What each number answers

For this cleaner, original length minus clean length equals removed characters because the current script deletes matched code points without inserting replacements. A count of zero means none of this tool's supported code points matched. It does not prove that the text contains no other invisible characters. The report gives totals by category, not the positions of changes; compare a copy of the original with the result before publishing.

Why the text can look unchanged

A public JavaScript question about a visually empty string with length one and another about a zero-width space in pasted code show the practical confusion: visual appearance and code-point count answer different questions. This example is a controlled test of this site's rules, not evidence that Claude placed those marks in any particular text.

Review meaningful characters before copying. In particular, a removed emoji joiner, Persian non-joiner, or em dash can alter the result. The cleaner does not rewrite sentences or predict AI-detector results.