Research and benchmark studies by the AI Aware team, led by Dr Tillman Weyde.
This report gives the full results behind the accuracy figures on our website. It covers the best result from each AI Aware detector in the 12 weeks to 7 November 2026: text, images and video, and audio. Alongside the headline figures, it shows how each detector performs by type of content, by AI model and
AI detector accuracy: full technical results for text, image, video and audio Read More »
On a benchmark of human and AI text across genres from fiction to medical writing, AI Aware was correct 98% of the time, the best result against Originality.ai, Winston AI, GPTZero, ZeroGPT, Ryne and Smodin. Over 99% of the texts it flags as AI are AI.
AI Aware beats major competitors in AI-generated text detection Read More »
Without any training on DeepSeek, AI Aware correctly identified 5,999 of 6,000 DeepSeek-v3 texts (99.983%). The results also show DeepSeek’s writing is close in style to GPT-3.5 and Grok-2.
AI Aware detects DeepSeek at first sight Read More »
