Why does my AI-generated text contain invisible characters?

AI models, editors, and copy-paste from web pages can silently insert zero-width characters, BOMs, variation selectors, and soft hyphens. These artifacts are invisible but can cause form validation errors, unexpected line breaks, and problems when pasting into editors, databases, or code.

Does NoAtMark remove AI watermarks?

No, and no tool can today. Claude's text watermark is a statistical signal in word choice — not a visible character — and C2PA lives in file metadata. NoAtMark cleans invisible Unicode characters; it doesn't claim to strip statistical watermarks or provenance.

Is the scanner private?

Yes. Everything runs locally in your browser. Your text is never uploaded to a server.

Free text hygiene for the AI age

Clean text, free and in-browser.

Q: What are zero-width characters?

Zero-width characters are Unicode characters that take up no visible space, such as the zero-width space (U+200B), zero-width joiner (U+200D) and zero-width non-joiner (U+200C). They are invisible in normal text but still occupy bytes and can break validation, searching, and formatting.

Paste any text — AI output, untrusted input, or a web copy — and instantly find every hidden Unicode and zero-width character. Every tool is free, private, and runs entirely in your browser.

🔒 100% private — in-browser ✅ Free — no sign-up 🔍 Deterministic — no false positives

0 chars

🔍

Ready to scan Paste text on the left, then hit Scan. Every invisible & zero-width character will be highlighted right here.

🔐

Private by designText never leaves your browser.

⚡

InstantScan on every keystroke.

🧹

One-click cleanStrip all hidden chars & copy.

🛡️

Zero uploadNo account, no tracking.

All tools

Clean · text

Invisible character remover

Strip zero-width trackers & hidden marks from pasted text. →

Clean · files

File cleaner

Scan then clean invisible characters in TXT, HTML, DOCX & more. →

Clean · formatting

LLM text formatter

Remove stray fences, blank lines & heading noise from AI output. →

Convert

Plain text extractor

HTML or rich text → clean plain text. →

Scan · security

Prompt injection scanner

Detect hidden text & injection patterns before pasting into an AI. →

Watermark

Invisible watermark

Add or verify a hidden, keyed watermark in your text. →

Scan · provenance

C2PA Checker

See whether an image carries Content Credentials. →

⭐ Flagship

Pre-Publish Inspector

Article health & GEO score before you publish. →

Developers

Text hygiene API

The same engines as a JSON API for your own pipelines. →

Security · CSV

CSV injection sanitizer

Escape = + - @ formula prefixes in CSV. →

Security · RAG

RAG sanitizer

Clean untrusted text before it hits your LLM. →

Audit

Submission auditor

Reveal hidden text in submitted work. →

Integrations — MCP, API, extensions & plugins

The same engines as an MCP server, a JSON API, browser extensions, and plugins for your CMS and writing tools. See all →

MCP

MCP server

AI agents clean text as they read files or search. →

API

JSON API

Sanitize, scan, and clean from your pipeline. →

Browser

Chrome / Firefox / VS Code

Clean automatically as you copy. →

WordPress

WordPress plugin

Auto-clean on save + bulk clean. →

Ghost

Ghost script

Bulk-clean via the Admin API. →

Notion

Notion script

Clean page text via the Notion API. →

Google Docs

Google Docs Apps Script

Add a NoAtMark menu to your doc. →

Clean text in three simple steps

Paste

Drop in AI output, a web copy, or a document — plain text only, nothing leaves your browser.

Scan

NoAtMark highlights every invisible & zero-width character and names each one (U+200B, ZWJ, BOM…).

Clean

One click strips the hidden artifacts and copies clean text — ready for editors, forms, code, or CMS.

What NoAtMark is — and isn't

✓ It cleans invisible characters

Zero-width spaces (U+200B), joiners, BOMs, direction marks, variation selectors — the hidden Unicode bytes that break validation, matching, and layout.

✗ It can't remove AI watermarks

Claude's text watermark is a statistical signal in word choice, not a character. C2PA lives in file metadata. Neither is something this tool — or any tool — strips today. Why →

Learn — the AI watermark era

Blog

All articles & guides

Invisible characters, AI watermarks, and text hygiene — explained. →

Claude's text watermark

What it is — and what can & can't be removed. →

Zero-width characters

U+200B, ZWJ, BOM — the invisible bytes. →

C2PA & Content Credentials

The signed-provenance standard, explained. →

AI watermark glossary

Every term in one place. →

Character reference

Zero-width space, BOM, and friends. →

Frequently Asked Questions

What are zero-width characters?

Zero-width characters are Unicode code points that occupy no visible space, like the zero-width space (U+200B), zero-width joiner (U+200D), and zero-width non-joiner (U+200C). They're invisible on screen but still count as characters, breaking validation, searching, and layout.

Why does AI-generated text have hidden characters?

AI models, text editors, and copy-paste from web pages can silently insert zero-width characters, BOMs, soft hyphens, and variation selectors. The output looks fine but carries invisible artifacts.

Does NoAtMark remove Claude's watermark?

No, and no tool honestly can today. Anthropic's text watermark is a statistical signal in word choice, not a visible character; C2PA lives in file metadata. NoAtMark removes invisible Unicode characters — it doesn't claim to strip statistical watermarks.

Is NoAtMark free?

Yes — every web tool is free forever, with no sign-up, no uploads, and no limits. The MCP server, API, and extensions are also free to use.

Can I clean invisible characters in bulk?

Yes — the WordPress plugin, Ghost script, and REST API all clean an entire site's content before publishing.