How accurate is grammarly ai detector?

Grammarly's AI detector, like most AI detection tools currently available, is not perfectly accurate and should be treated as a probabilistic indicator rather than a definitive verdict. It analyzes patterns in writing, like sentence structure, word choice, and predictability, to estimate the likelihood that text was generated by AI, but it can produce both false positives, flagging genuinely human-written text as AI-generated, and false negatives, missing AI text that's been edited or paraphrased. Accuracy tends to drop further when the text has been lightly rewritten after being AI-generated, since that editing process disrupts the patterns the detector is trained to recognize. Independent testing of AI detectors, including Grammarly's, has generally found that no tool on the market achieves consistently high accuracy across all types of writing and all AI models, and results can vary depending on the length and style of the text being analyzed. Because of these limitations, most experts recommend against using any single AI detector as the sole basis for high-stakes decisions like academic discipline, and instead suggest using it as one signal among several, combined with human judgment.