AI Struggles to Mimic Human Imperfection, Remains easily Detectable
New research reveals a surprising hurdle in AI progress: convincingly replicating human interaction is proving far more challenging than demonstrating intelligence. A recent study, detailed in a paper published on arXiv (https://arxiv.org/abs/2511.04195), shows that even advanced AI models are frequently identified as non-human in a modified Turing test, with an accuracy rate of 80%.
Researchers from the University of Zurich, University of Amsterdam, Duke University, and New York University collaborated on the study, which focused on how well AI could mimic human language across social media platforms X, Bluesky, and Reddit. They developed a ”computational Turing test” utilizing classification systems and linguistic analysis to assess the human-likeness of AI-generated text.
Nine AI models were put to the test: Llama 3.1 8B, Llama 3.1 8B Instruct, Llama 3.1 70B, Mistral 7B v0.1,Mistral 7B Instruct v0.2,Qwen 2.5 7B Instruct, Gemma 3 4B Instruct, deepseek-R1-Distill-Llama-8B, and Apertus-8B-2509. The results consistently showed that AI-generated content was identified as such in 70 to 80% of cases.
The key finding? AI consistently fails to replicate the nuances of human negativity and spontaneity. Researchers observed that the models tended to be overly positive and lacked the informal, sometimes unpleasant, communication style common in human interactions. Even when explicitly instructed to emulate existing online conversations, the AI’s output remained noticeably artificial.
“It is indeed striking that even if the AI is instructed to answer as much as possible as other peopel have done, it is even easier to determine that it is AI,” the study indicates. Furthermore, the research suggests that increasing the sophistication of the AI system doesn’t automatically translate to more human-like output.
While the ease with which AI can be detected is reassuring to some, the study also raises a point about human online behavior. The researchers suggest that the very characteristic making AI stand out – a lack of “unpleasantness” – highlights a perhaps concerning trend in social media communication, prompting reflection on how users interact online.
