Paste any text and see the likelihood it was written by AI or a human.
Copy the text you want to check and drop it in the box.
The model runs locally in your browser β nothing is uploaded.
An estimated probability of human vs. AI-generated writing.
A quick look at what actually happens between pasting your text and seeing a result.
Your text is split into tokens (words and sub-word pieces) the model can work with, and trimmed to the model's maximum input length if it's very long.
The model looks at statistical patterns in how the words are put together β things like word predictability, sentence rhythm, and repetition β that tend to differ between human writing and machine-generated text.
Two signals matter most: perplexity (how "surprising" the word choices are to a language model β AI text tends to pick more predictable words) and burstiness (how much sentence length and complexity vary β human writing is usually less uniform).
A trained classifier weighs all these signals together and outputs a confidence score for each label β "Real" (human-written) or "Fake" (AI-generated).
The score is converted into the human vs. AI percentages you see on screen, along with a plain verdict and the honesty disclaimer above.