AI Fake News Rated More Credible Than Human Texts

AI-generated disinformation is frequently perceived as more credible and informative than human-written text, according to new research from the NxtGenFake project at the University of Oslo ^[1][2]. In a study involving American participants, researchers found that readers rated AI-produced fake news higher on parameters of trustworthiness and informativeness when the source was not disclosed ^[2][3]. Unlike human authors who often use emotional triggers, AI models utilize a consistent and subtle set of persuasive techniques ^[3]. The study noted that large language models often employ generic "Appeals to Authority," using phrases such as "experts believe" or "according to researchers" to mimic the authoritative tone of legitimate journalism ^[1][2]. This professional formatting allows misinformation to slip into genres that audiences instinctively trust ^[1]. Lead linguist Silje Susanne Alvestad warned that these findings suggest AI disinformation may be significantly harder to detect than previous iterations ^[2]. The NxtGenFake project, funded by the Research Council of Norway, plans to continue its work through 2029 to develop linguistic tools capable of identifying these artificial patterns ^[1][2].