AI Text Cleaner
Paste text from ChatGPT, Claude, Gemini or Perplexity and remove what gives it away: hidden characters, em dashes, citation markers and stray Markdown — without changing a word.
Free, no signup — cleaned in your browser, never uploaded.
Certainly! Here’s…and
Let me know…won’t belong in your clean text.
Changed 1 odd space, 3 emoji bullets, 2 dashes → commas, 5 smart quotes, 1 ellipsis Removed 1 hidden character, 3 citation markers, 10 Markdown symbols Flagged 3 AI phrases (not removed)
How to Clean AI Text
- Paste the answer. Copy it with the chat’s copy button or select it by hand — from ChatGPT, Claude, Gemini, Perplexity or any other assistant — and paste it on the left.
- Read the chips. Each rule shows how many things it found in your text. Every rule is on except the ones you switch off; hover a chip to see exactly what it does.
- Choose what an em dash becomes: a comma (the default), a hyphen, a spaced hyphen, nothing, or keep it. Number ranges such as 2020–2024 always become 2020-2024.
- Check the Diff. It shows every change in place: removed text struck through in red, added text in green, and hidden characters as named tags such as
ZWSPandNBSP. - Copy clean text — or open the Markdown tab to keep the headings and lists, and use Send to for a PDF, a Word document, a LinkedIn post or rich text.
What Gets Cleaned
Each rule is a chip above the panes, with a live count of what it found. All of them are on until you switch one off.
| Rule | What it does | Example |
|---|---|---|
| Invisible characters | Zero-width spaces and joiners, byte-order marks, stray direction marks and hidden tag characters. Emoji sequences and right-to-left text keep the ones they need. | research (with a zero-width space) → research |
| Non-breaking & odd spaces | Non-breaking, narrow, thin and other special spaces become ordinary spaces, and non-breaking hyphens become ordinary hyphens. | 40 minutes (non-breaking space) → 40 minutes |
| Citation markers | Source markers such as [1], 【4†source】, citeturn0search3, (source) and [cite: 2], and utm_ tracking parameters on links. | Sales rose 4%【4:0†source】. → Sales rose 4%. |
| Emoji bullets | Emoji used as bullets (✅ 🚀 👉) become plain list items; emoji in front of headings are removed. | ✅ Faster charging (in a list) → - Faster charging |
| Em & en dashes | Choose what an em dash becomes. Number ranges such as 2020–2024 become 2020-2024. | Fast — and free → Fast, and free |
| Smart quotes | Curly quotes and apostrophes become straight ones. | “It’s done” → "It's done" |
| Ellipsis character | The single … character becomes three dots. | Wait… → Wait... |
| Markdown symbols | Plain text removes **, #, backticks and other Markdown syntax; lists, links and tables keep their structure. | **Key point:** done → Key point: done |
| Trailing spaces | Spaces and tabs at the end of lines. | End of line␠␠␠ → End of line |
| Extra blank lines | Runs of empty lines become a single blank line. | Three empty lines in a row → One empty line |
| “As an AI…” phrases | Highlights assistant boilerplate, openers and sign-offs for you to review. Nothing is removed. | As an AI language model, … → Highlighted for you — never removed |
What the Cleaner Leaves Alone
- Your words. No rule rewrites, reorders or deletes a sentence. Chatbot phrases are highlighted for you to decide on; the “Certainly! Here’s…” opener and “Let me know if…” sign-off can be removed with one click, and that is undoable.
- Code. Dashes, quotes, ellipses and citation-like brackets inside code blocks and inline code are not touched; only hidden characters and odd spaces are fixed there, because those are the ones that break code.
- Emoji and other scripts. Emoji in a sentence stay. Joiners inside emoji like 👩💻 and flags stay, and so do the direction marks right-to-left text needs.
- Your own references. Real links, footnotes such as
[^1], task boxes and reference definitions are kept; only the markers chat assistants add are removed.
Plain text or Markdown?
The Clean text tab is plain text for email, forms and messages. The Markdown tab applies the same cleaning but keeps the Markdown, which is what you want before a converter: Send to hands that version to Markdown to PDF, Markdown to Word and the LinkedIn Text Formatter.
AI Text Cleaner vs AI Detector vs Humanizer
A cleaner (this tool) removes the characters and formatting that come along when you copy an AI answer: hidden Unicode, typographic dashes and quotes, citation debris, Markdown symbols. It never rewrites a sentence.
An AI detector estimates whether text was machine-written from its wording. Cleaning does not change the wording, so it is not a way around a detector, and it is not meant to be one.
A humanizer rewrites sentences to change how they read. That can change your meaning; review anything it gives you. If you only need the text to look clean when you paste it, a cleaner is the safer tool.
About “AI watermark removers”
Some sites promise to strip an “AI watermark”. Character cleaning can’t do that: statistical watermarks such as Google’s SynthID live in word choice, not in hidden characters. What this tool removes are the stray characters that copying leaves behind — and it shows you every one in the Diff.
Frequently Asked Questions
How do I remove em dashes from ChatGPT text?
Paste the text here. The Em dashes rule is on by default and turns each em dash into a comma; open its menu to use a hyphen, a spaced hyphen or nothing instead, or to keep them. En dashes in number ranges (2020–2024) become plain hyphens, a dash that opens a line becomes a list bullet, and dashes inside code are left as they are.
How do I get rid of the asterisks (**) and # in ChatGPT answers?
They are Markdown: ChatGPT writes **bold** and ## headings, and they show up as symbols when you paste into a plain text box. With the Markdown rule on, the clean text has none of them, while lists, links and tables keep their structure. If you want the bold and the headings to survive instead, send the text to Markdown to Rich Text and paste it into Google Docs, Word or Gmail as formatted text.
Does the cleaner change my words?
No. Every rule works on characters, punctuation and formatting: hidden characters, spaces, dashes, quotes, citation markers and Markdown symbols. “As an AI…” phrases and chatbot sign-offs are only highlighted for you to review. The Diff view shows every single change, so you can check.
What are hidden characters, and why are they in AI text?
Characters that take no visible space: zero-width spaces and joiners, byte-order marks, direction marks, non-breaking spaces, and the private-use glyphs ChatGPT wraps around its citations. They come along when you copy from a chat window and can break search, code, word counts and spell checkers. The cleaner removes them, but keeps the joiners that hold emoji like 👩💻 together and the direction marks right-to-left text needs.
Will cleaning make text undetectable by AI detectors?
No, and it does not try to. The cleaner fixes characters and formatting; the wording, which is what readers and detectors judge, stays exactly as it was. Statistical watermarks such as Google’s SynthID live in word choice, not in hidden characters, so no character cleaner can remove them.
Can I keep the Markdown formatting?
Yes. Open the Markdown tab to get the same cleaning with the Markdown kept, or switch the Markdown rule off. Send to always hands on the cleaned Markdown, so headings and lists survive in a PDF, a Word document or a LinkedIn post.
Which assistants does it handle?
Anything you can paste. It knows the citation formats of ChatGPT (【4:0†source】, citeturn0search3, utm_source=chatgpt.com links), Gemini ([cite: 2]) and Perplexity ([1]), and the typographic habits all of them share: em dashes, curly quotes, the ellipsis character and emoji bullets.
Is my text uploaded?
No. Cleaning, counting and the diff all run in your browser, and nothing you paste is sent to a server. The page’s analytics record that the tool was used, never the text.