AI vs. the Fakes: Inside the 2025 Race to Spot ChatGPT, Deepfakes and More
GPTZero, created by a Princeton student in early 2023, was among the first famous AI text detectors and highlights AI-written sentences with color-coding and an “AI probability” score. OpenAI discontinued its AI-written text detector in 2023 after finding a low rate of accuracy. In a Washington Post test, Turnitin’s AI detector got over half the essays at least partly wrong and flagged 8% of a student’s original essay as AI. The European Union’s AI Act of 2024 requires AI-generated content to be disclosed, with providers mandated to watermark or clearly indicate AI origin starting in 2026. Google announced SynthID in