Artificial intelligence has become an inevitable part of contemporary life and has significantly influenced the field of translation and translation studies. While AI has been widely employed in screen content production, it has also increasingly been integrated into audiovisual translation.
Naturally, many of the questions traditionally raised about human translation are now being reconsidered within translation studies, with a renewed focus on machine translation and, more specifically, on AI-assisted translation. The translator’s cultural background (Cordonnier, 2002: 38), the close relationship between language and perception (Cadiot and Visetti, 2001: 1), and the importance of continuity (Visetti, 2004: 39) raise new theoretical and methodological questions when translation is mediated by AI systems that lack lived cultural experience and rely instead on training data and computational processes.
In the context of AI-assisted translation, “translation is reconceptualized as an emergent phenomenon produced through the interaction of heterogeneous agencies, including human translators, algorithms, training corpora, interfaces, platforms, and end-users” (Alowedi et al., 2026: 3718). Unlike human translators, AI-generated outputs are shaped by training data, model architectures, and optimization processes rather than lived cultural experience.
This study analyzes English-to-Persian subtitles produced through an AI-only workflow in order to investigate their effects on cultural meaning, stylistic richness, and interpretive openness. Through a qualitative comparative analysis of six purposively selected dialogue sequences from two English-language horror films, Get Out (2017) and The Babadook (2014), it compares professional human-produced and AI-generated subtitles. The sequences are selected for their culturally embedded expressions, metaphorical formulations, stylistically marked language, and/or interpretive ambiguity. The AI subtitles are generated using ChatGPT (GPT-5.5), without human post-editing, and are compared with human-produced Persian subtitles.
The analysis focuses on three interconnected dimensions: cultural meaning, metaphorical and stylistic richness, and interpretive openness. The study hypothesizes that AI-generated subtitles may tend to (1) standardize or normalize culturally embedded expressions, (2) favour more explicit or conventional renderings over metaphorically and stylistically marked formulations, and (3) reduce interpretive openness by resolving ambiguity. The study thus seeks to determine whether AI-assisted subtitling tends to reduce or increase cultural and interpretive complexity in favour of greater standardization and efficiency.
Bahareh Ghanadzadeh Yazdi holds a PhD in Translation Studies (Traductologie) and Comparative Literature (littérature comparée). She obtained her doctorate from the University of Artois in France, where she defended her dissertation in 2019. The title of her thesis, written in French, is Traduire en français et en persan les métaphores de The Hunger Games de Suzanne Collins : réflexions théoriques et problématiques traductologiques. Bahareh Ghanadzadeh Yazdi has worked as a professional translator for over 25 years and has more than 50 published translations into Persian. Her main areas of expertise include young adult literature and subtitling. Her doctoral dissertation was published as a book in 2022 by Les Classiques Garnier in France. She worked as a lecturer and researcher at the University of Évry (Paris) from 2021 to 2023. Since 2023, she has been working as an administrative staff member in research laboratories.
