Field notes on AI detection: where detectors disagree with each other, how often they flag writing that a person actually wrote, and what happens when three engines read the same text. Most of it comes out of our own scan data, and the sample size is always attached to the number.
Does Grammarly detect AI-generated text? We tested it. Grammarly added AI detection in 2024, but it uses a single engine. See how it compares to multi-engine detection.
Turnitin detects AI in English, Spanish and Japanese only — and the Washington Post measured a 50% false-positive rate. See the real numbers, and get a 3-engine second opinion before you are accused.
Show readers your content is human-verified. Embed the free OmniDetect Verification Badge on your blog or website in under 2 minutes — no coding skills required.
The same text gets 12% on GPTZero and 67% on Originality.ai. This is not a bug — it is how AI detection works. We explain why detectors disagree, what the scores actually measure, and how multi-engine consensus solves the problem.