When the algorithm learns to tell stories

a comparative study of telenovela scripts generated by artificial intelligence

Abstract

The expansion of generative artificial intelligence has brought about significant transformations in audiovisual creative processes. This article examines the production of telenovela scripts generated by generative AI systems, investigating their limits, recurrences, and narrative potentialities. The study is guided by the hypothesis that, although these technologies are capable of producing structurally coherent texts aligned with genre conventions, their outputs tend toward algorithmic standardization, the reproduction of clichés, and constraints in the construction of complex and culturally situated affects. The research adopts a qualitative and exploratory approach, based on the experimental generation of scripts using the same prompt across three distinct AI platforms (ChatGPT, Gemini, and DeepSeek), followed by a comparative analysis focusing on textual coherence, dramatic structure, character construction, and strategies for producing affect. The theoretical framework draws on contributions from narrative studies, digital culture, and artificial intelligence, engaging authors who conceptualize AI as a cultural device embedded in regimes of symbolic mediation and algorithmic predictability. The findings reveal significant structural convergences among the scripts produced and analyzed. The article concludes that generative AI operates less as an autonomous creative agent and more as a system of probabilistic recombination of consolidated narrative forms, reinforcing the need for critical, ethical, and interdisciplinary approaches to its use in contemporary audiovisual production.