Paste text to scan for Unicode characters commonly produced by AI — em dashes, smart quotes, guillemets «»‹›, ellipsis, and invisible glyphs. Use Fix to replace them with plain keyboard characters.
| Char | Unicode | Name | Replaces with | Found | AI signal |
|---|
Large language models (LLMs) like ChatGPT, Claude, and Gemini tend to produce certain Unicode characters that are rare in normal human typing. These act as subtle fingerprints of AI-generated text.
This tool scans pasted text for those characters, highlights them, and lets you replace them all with plain ASCII equivalents in one click.
LLMs learn from text scraped from the web and books, much of which was typeset with proper Unicode punctuation. The models reproduce these stylistic choices even when generating plain prose — producing em dashes instead of hyphens, curly quotes instead of straight ones, and so on.
Invisible characters (zero-width spaces, joiners) sometimes appear due to copy-paste artifacts in training data or as tokenization quirks.
This is a heuristic, not a classifier. A document full of em dashes from a human writer using a Mac keyboard (Option+Shift+Hyphen) will trigger false positives. Conversely, AI text with no special characters won't be flagged at all. Use results as one signal among many.