Blog
Technical guides on how hidden characters get into AI-generated text, how detection works, and how ChatGPT, Claude and Gemini output differs.
Why Does ChatGPT Use So Many Em Dashes? (And How to Remove Them)
Em dashes come from training data and preference tuning, not a watermark. Why ChatGPT overuses them, how to stop it, and how to remove them from text you have.
Read more →Does ChatGPT Watermark Text? No — Here's What Actually Gets Copied
No, ChatGPT does not add a real watermark. Here is what the HTML attributes and invisible characters people mistake for one actually are.
Read more →Zero-Width Space in Code: Why It Looks Right but Fails
A zero-width space in a variable or config key compiles fine to the eye but fails at runtime. Which characters cause it and how to find them.
Read more →Invisible Unicode Characters: The Complete Reference List
A reference table of invisible Unicode characters — code points, categories, and whether each one is safe to strip from your text.
Read more →How to Clean ChatGPT Output Before Pasting Anywhere
What rides along when you copy from ChatGPT, how ten common destinations handle it, and a simple 3-step cleanup routine.
Read more →How to Detect Claude AI Text: Markers and Technical Fingerprints
Anthropic's Claude has a different clipboard footprint than ChatGPT. Learn what Claude's markers look like, how they differ, and how to detect them.
Read more →How to Detect Gemini AI Text: Google's AI Fingerprints Explained
Google's Gemini leaves unique technical markers in copied text. Learn what Gemini's fingerprints look like and how they compare to ChatGPT and Claude.
Read more →