AI models like ChatGPT, Claude, and Gemini often insert invisible Unicode characters into generated text and code — curly quotes, em dashes, non-breaking spaces, zero-width joiners, and more. These break code, cause encoding issues, and produce invisible bugs. Paste your text below to clean it instantly.
What does this tool fix?
Typography characters
- Curly quotes (" " ' ') → straight quotes
- Em dash (—) → removed
- En dash (–) → hyphen
- Ellipsis (…) → three dots
- Bullet (•) → hyphen
Invisible characters
- Non-breaking spaces (U+00A0, U+202F)
- Zero-width spaces (U+200B, U+200C, U+200D)
- Directional marks (LTR/RTL overrides)
- Variation selectors (U+FE00–U+FE0F)
- BOM (U+FEFF) and other invisible controls
Why do AI models add these characters?
Large language models are trained on diverse text sources including published books, web pages, and formatted documents. These sources use typographically correct characters like curly quotes and em dashes. When you paste AI-generated text into code, terminals, config files, or APIs, these characters cause silent failures — a curly quote in a Python string, a non-breaking space in YAML indentation, or a zero-width character in a variable name.