“Once upon a time, there was a civil war”: Automatic adaptation of museum object descriptions for different age groups with GenAI

Abstract Adapting texts for different age groups can be desirable but costly and therefore often unfeasible. Large language models (LLMs), with their text manipulation capabilities, may offer an automated solution. This article explores the use of LLMs for text simplification for children of different age groups. To this end, we submitted approximately twenty prompt variations for adapting texts to different age groups to four LLM-based chatbots (Copilot, ChatGPT, DeepSeek, and Claude) and compared the resulting outputs, using a specially developed evaluation framework as a point of departure. The source text (ST) selected for the experiment was an English translation of a Finnish explanatory text for an exhibited item from a Finnish Civil War-themed exhibition at the Vapriikki Museum in Tampere, Finland. Our results suggest that current models, while modestly capable of performing this task for older teens, struggle to produce adequate adaptations for younger age groups. LLMs show great promise but they still require substantial human revision to yield satisfactory results for the youngest audiences. Our main contribution is the examination of both the possibilities that LLMs may offer for intralingual translation in museum contexts and the specific challenges and areas of concern that may arise for potential post-editors working with LLM-generated adaptations.