Scan the Unicode string
The cleaner iterates by Unicode code point rather than by visible glyph. It groups suspicious characters under readable names such as zero-width space, non-breaking space, or bidirectional override.
Find and remove hidden characters from text
Reveal zero-width spaces, non-breaking spaces, byte order marks, bidirectional controls, and other invisible Unicode characters before copying clean text.
Invisible does not always mean useless
- Zero-width spaces, soft hyphens, byte order marks, and word joiners can interrupt matching, validation, wrapping, or identifiers.
- Bidirectional overrides and isolates can change display order and create confusing text even though the stored characters remain unchanged.
- Non-breaking, narrow, thin, figure, and ideographic spaces may look alike while behaving differently in wrapping and comparison.
- Joiners, non-joiners, variation selectors, and direction marks can be meaningful in emoji and human writing systems.
Clean the text without losing control.
Apply a conservative pass
Essential mode removes common clipboard artifacts, control characters, and unsafe bidirectional controls while preserving emoji joiners, non-joiners, variation selectors, and basic direction marks.
Escalate only when needed
Deep clean removes a broader set of format characters, annotation controls, tags, fillers, and variation selectors. Use it for strict technical input only, then review language and emoji carefully.
Verify the destination behavior
Copy the clean result into the actual field and test wrapping, search, validation, cursor movement, export, and downstream processing. A clean-looking preview is not the only acceptance test.
Know what the cleanup can prove.
Readable inspection
See character names, code points, and counts rather than guessing whether a blank area contains an ordinary space.
Deterministic cleanup
The same input and preset produce the same result. No model rewrites the wording, tone, facts, or structure behind your back.
Safer defaults
The standard modes protect characters that may carry meaning. The strongest removal remains an explicit, reversible choice.
Keep formatting and provenance separate.
Removing hidden Unicode characters does not remove provider-side logs, provenance held outside the string, signed metadata in a file container, or statistical patterns encoded through word choice. It also does not establish who wrote the text.
Common questions, answered plainly.
What is a zero-width space?
Unicode U+200B is an invisible break opportunity. It can help line breaking in some writing systems, but an unexpected instance can disrupt search, identifiers, and copy-paste behavior.
Are bidirectional controls dangerous?
They have legitimate uses, but overrides and isolates can also make displayed order differ from stored order. The cleaner removes the riskiest controls and reports them.
Why not remove every format character automatically?
Some format characters shape scripts, emoji, and text direction. Automatic blanket removal can damage valid text, so the broadest pass is a separate Deep clean mode.
Paste the text. See every fix.
Use the same canonical cleaner for hidden characters, unusual spaces, Markdown, HTML, smart punctuation, and clean plain-text copy.
Make copied text useful without making false promises.
Clean what is present, review what changed, and preserve any disclosure or editorial record the work still needs.
Open the AI Text Cleaner