Claude Text Cleaner

Accent and normalization guide · October 31, 2026

Does Claude Text Cleaner Normalize Accented Text?

No. The browser cleaner does not turn e + U+0301 COMBINING ACUTE ACCENT into the single code point é, or remove the accent. A live test left both inputs unchanged: é stayed one code point, while é stayed two. Each report showed zero removals. The two strings can look alike while remaining different sequences.

Try the browser cleaner

Code-point labels show the difference even when a font renders both forms similarly. The live test used the actual U+0301 mark after e.

Why does the cleaner leave the two forms different?

The cleaning script checks each input code point against seven fixed removal categories, then copies every unmatched code point to the result. U+0301 and U+00E9 match none of its categories. The script does not call a Unicode normalization function. Its zero-removal count describes only those removal rules; it does not mean two strings have the same underlying code points.

Unicode's normalization FAQ explains that a precomposed letter and a base letter plus combining mark can be canonically equivalent while encoded differently. An English-language Stack Overflow question gives this exact é versus e + U+0301 comparison in JavaScript. That discussion is about string equality, not a report about this cleaner.

How to check a mismatch safely

  1. Keep the original text. Paste a copy into the cleaner and read its before/after lengths. Copy the full é example here if testing; typing a keyboard é may insert the one-code-point form instead.
  2. Inspect code points. In a development console, Array.from(text, c => c.codePointAt(0).toString(16)) shows the sequence. The cleaned result retains the same sequence in the examples above.
  3. Normalize only where your workflow needs it. If exact matching is the problem, decide on a Unicode normalization form in the destination application and apply it consistently to the values being compared. This cleaner is not a normalization tool; changing text without checking language and storage requirements may be inappropriate.

The live browser test on October 31, 2026 returned lengths 1→1 for é and 2→2 for é, with zero removed characters and zero changed categories in both cases. For the different U+034F mark that can interrupt normalization, see the Combining Grapheme Joiner guide. This cleaner cannot establish text origin, detect a watermark, or predict an AI-detector result.