Invisible Text Remover
Find and remove hidden Unicode characters such as zero-width spaces, em dashes, smart quotes, and invisible markers, in text copied from ChatGPT, Claude, Gemini, websites, or documents. Every scan and cleanup runs locally in your browser; your text never leaves your device.
Updated
We hunt for characters like
Paste text above and every hidden character will light up here as a badge.
What Are Invisible Characters?
Invisible characters are Unicode code points that take up no visual space or look like something they are not. The most common ones, zero-width spaces (U+200B), zero-width joiners (U+200D), and left-to-right marks (U+200E), are completely invisible to the eye but still exist in your text. They get copied into your content when you paste from ChatGPT, Claude, Gemini, websites, PDFs, or word processors, and once there they can break string matching, break URLs, corrupt JSON and CSV files, inflate word counts, and cause "why won't this paste correctly?" problems. Closely related are visible special characters that AI models overuse, such as em dashes (U+2014), smart quotes, and ellipses, which make text look machine-generated and can trip up code editors and data pipelines.
How to Remove Hidden Characters from AI Text
- Paste your text into the input pane: from ChatGPT, Claude, Gemini, a website, or a document. Everything stays in your browser.
- Review the analysis panel: it counts invisible and visible special characters and lists each detected type with its Unicode name and purpose.
- Reveal hidden characters: the reveal section shows exactly where each hidden character sits, rendered as badges. Use the toggles to focus on invisible characters, unicode spaces, punctuation, or line breaks.
- Clean the text: click Remove invisible to strip zero-width characters, Fix quotes & dashes to convert smart punctuation to plain ASCII, or Clean all for both in one step.
- Copy the result: click Copy to grab the cleaned text for your document, CMS, or codebase.
Common Hidden Characters in AI-Generated Text
| Character | Unicode | What it is | What we do |
|---|---|---|---|
| Zero-Width Space | U+200B | Invisible space with no width that breaks matching, URLs, and word counts | Remove |
| Zero-Width Joiner / Non-Joiner | U+200D / U+200C | Invisible glue characters used in emoji and script shaping | Remove |
| Left/Right-To-Right Mark | U+200E / U+200F | Invisible text-direction markers | Remove |
| Soft Hyphen | U+00AD | Invisible line-break hint | Remove |
| Variation Selectors | U+FE00–U+FE0F | Invisible glyph selectors, common near emoji | Remove |
| Byte Order Mark | U+FEFF | Invisible file/copy-paste artifact that breaks parsing | Remove |
| No-Break Space | U+00A0 | Looks like a space but prevents line breaks | → space |
| Em Dash | U+2014 | The long dash ChatGPT inserts constantly | → "-" |
| Smart Quotes | U+2018–U+201D | Curly quotes that break code and CSV | → straight |
| Ellipsis | U+2026 | Single-character "…" replacing three dots | → "..." |
| Line / Paragraph Separator | U+2028 / U+2029 | Invisible structural separators | → newline |
| Hangul / Braille fillers | U+3164 / U+2800 | Blank-looking characters used as fake spaces | → space |
Does Removing Invisible Characters Bypass AI Detectors?
No. Removing hidden characters does not make AI-generated text undetectable. AI-detection vendors have published tests showing that adding or stripping invisible Unicode characters leaves detection results essentially unchanged. Detectors analyse word choice, repetition, perplexity, and sentence structure, not hidden code points. This tool exists to clean up formatting and fix real bugs (broken URLs, bad CSV, failed string matching). If you are looking for an "AI watermark remover" in the hope of evading detection, it will not work, and nothing reliably does.
Why Do LLMs Insert Hidden Characters?
ChatGPT, Claude, and other large language models frequently produce characters you cannot see or type. There are three main reasons: they train heavily on professionally edited text where em dashes, smart quotes, and thin spaces are standard typography; those characters make output look more formal and authoritative; and unlike a human, a model never touches a keyboard, so there is no extra "cost" for emitting a non-keyboard character. OpenAI has described the behavior as a quirk of large-scale reinforcement learning rather than an intentional watermark.
Are Invisible Characters Dangerous?
Mostly they are a nuisance, but they can be actively harmful. Hidden characters change how software wraps lines, splits words, hashes strings, and matches text: two passages that look identical can fail an equality check or a search. They can hide malicious instructions in code or prompts, and security researchers have documented Unicode exploits that smuggle extra instructions into chatbot prompts and make models behave unexpectedly. Attackers can also bury invisible threats in code so human reviewers see nothing unusual. When you receive text from an untrusted source, running it through this cleaner is a cheap safety step.
Frequently Asked Questions
What is a zero-width space?
A zero-width space (U+200B) is a Unicode character that occupies no visible space but still counts as a character. It is the most common hidden character in text copied from ChatGPT and websites, and it can break string matching, URLs, word counts, and JSON or CSV parsing.
How do I remove hidden characters from ChatGPT text?
Paste the text into the input pane, review the detected characters in the analysis panel, then click Remove invisible to strip zero-width spaces, joiners, and markers, or Clean all to also convert smart quotes and em dashes to plain ASCII. Everything happens locally in your browser.
Can you remove an AI watermark from text?
This tool removes invisible Unicode characters, which are sometimes loosely called an "AI watermark". Invisible characters are not OpenAI's watermarking mechanism. They are a known output quirk. Cleaning them fixes formatting but does not make AI-generated content undetectable.
Will removing invisible characters make text undetectable by AI detectors?
No. Tests by detection vendors show that adding or removing invisible characters barely changes detection results. AI detectors look at writing patterns and structure, not hidden code points. Use this tool to clean your text, not to evade detection.
Is there a free tool to remove hidden characters from AI text?
Yes, this page is exactly that. Paste text from ChatGPT, Claude, or Gemini into the input pane and it instantly counts and reveals hidden characters: zero-width spaces, joiners, direction marks, and variation selectors. Click Remove invisible or Clean all to strip them, then copy the result. Everything runs locally in your browser.
How do I get rid of invisible characters in my text?
Paste the text into this tool, review the detected characters in the analysis panel, then click Remove invisible to delete zero-width spaces and markers, or Clean all to also convert smart quotes, em dashes, and unicode spaces to plain ASCII. The cleaned text is ready to copy in one click.