The shape of AI slop
By now, most people who read these posts will be painfully familiar with the signs of AI-generated text.
Recent research by a team led by Jenna Russell from the University of Maryland has shed light on the notable differences between AI-generated and human-written stories. In an experiment involving five large language models (LLMs) such as Claude, DeepSeek, Gemini, GPT, and Kimi, along with human authors, the researchers classified stories based on 304 criteria.
The study found that AI stories are distinguishable from human stories in several aspects. First, AI stories tend to over-explain their themes, with 77% of AI-generated stories explicitly stating the theme compared to 52% of human stories. Secondly, AI stories are more blunt in expressing emotions, frequently mentioning physical sensations and metaphors related to bodily experiences (81% of AI stories vs. 38% of human stories).
Additionally, AI stories often use the setting as a metaphor for a character's inner state more frequently than human stories.
Another key difference lies in the structure of AI stories. They tend to be more linear and causal, with fewer subplots and ambiguous endings. In fact, 79% of AI stories do not include any subplots, whereas only 57% of human stories lack subplots. Moreover, human stories frequently employ time jumps and disjoint causal chains, contributing to their complexity.
In summary, AI-generated stories are simpler, less convoluted, and less diverse compared to human stories. The findings from this study highlight the unique narrative features that human authors instinctively incorporate to create interesting and complex stories, setting them apart from the more monotonous and formulaic output of AI models.
Written by urgent.news from Klement on Investing's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
