Invisible Text Detector
Detect and remove invisible Unicode characters from text
Detect and remove invisible Unicode characters, zero-width spaces, BiDi controls, and hidden steganography from text by ChatGPT. Find AI-injected artifacts, Trojan Source attacks, and prompt injection payloads. Free online invisible text detector with download.
ZWSP, ZWNJ, ZWJ, Word Joiner, BOM, Soft Hyphen. The foundation of Unicode steganography — they carry data while rendering invisible.
12 directional formatting characters (LRM, RLM, LRO, RLO, etc.) used in Trojan Source attacks (CVE-2021-42574) to make code display differently from execution.
17 space variants (NBSP, en/em spaces, thin/hair spaces, NNBSP). OpenAI o3/o4-mini inject Narrow No-Break Space (U+202F) in outputs.
Curly quotes, em dashes, ellipsis, and smart apostrophes. AI models inject these typographic characters instead of ASCII equivalents — common in ChatGPT, Claude, and Gemini outputs.
256 variation selectors can encode arbitrary binary data. Tag characters (U+E0001-E007F) mirror ASCII invisibly — used for prompt injection attacks.