What it does
You can label examples and tune the threshold, but stylistic signals cannot prove who or what wrote a post.
Playable demos & bots
Watch posts suspected of AI writing turn red or fold away on X, LinkedIn, and Reddit as you scroll.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
You can label examples and tune the threshold, but stylistic signals cannot prove who or what wrote a post.
A developer can borrow this method to score writing styles in your own project. They can set up code to send text blocks and specific questions to TypeSafe. This outside AI service answers each question with a probability. Your app then combines these answers to guess if the text is automated.
You will need an access key that connects the app to TypeSafe. Your developer can adjust the scoring weights for different questions. The software ignores posts under five words because there is too little text to judge. Remember that writing style alone cannot prove who actually wrote a message.