Keywords
Abstract
As large language models (LLMs) become more common in translation work, prompt engineering has emerged as a key factor influencing the quality of AI-generated literary translations.
However, while much of the existing research has focused on comparing human and machine translations, the specific effects of different prompting strategies on literary texts have received limited attention.
This study explores how prompt design affects translation quality in a literary context.
Using a comparative qualitative approach, the analysis examines selected passages from contemporary English fiction translated under three conditions: a basic instruction, a style-focused prompt, and a literarily oriented prompt.
Translations were evaluated according to semantic accuracy, stylistic preservation, metaphor rendering, and narrative coherence. The results indicate that prompt design has a clear and meaningful impact on translation outcomes.
Basic prompts tend to produce fluent but somewhat simplified versions, while literary prompts are more successful in retaining figurative language, stylistic nuances, and emotional depth.
At the same time, overly elaborate prompts can introduce unnecessary lexical complexity, which may reduce readability.
These findings suggest that prompt engineering should be regarded as an essential component of AI-assisted literary translation, rather than a mere technical detail.
The study contributes to translation studies by proposing prompt design as a relevant analytical variable and offers practical recommendations for translators and researchers working with large language models.