PIRD — Paraphrase-Robust AI-Text Detector

PIRD is a paraphrase-robust, cross-domain detector for AI-generated text. Unlike perplexity-based detectors, it stays reliable when AI text is paraphrased to evade detection. Paste a passage (≥ 20 words) to get a calibrated probability that it was AI-generated.

Research demo — predictions are probabilistic and not infallible; do not use as sole evidence of misconduct.