The aim of this paper is to investigate whether a large language model (ChatGPT with GPT 3.5) can be used to generate learner prompts as part of the development of a new application, Cén Scéal?, within the Irish (Gaeilge) iCALL platform, An Scéalaí. In this work an evaluation of the model’s accuracy in generating lists of words within various POS tags is presented, which enables an interrogation of the model’s ontology of the lexicon and of morphological rules. It is found that the model hallu- cinates lexical items, mistranslates words, and makes mistakes with grammatical gender. The consequences of the model’s error rates in these regards are discussed with respect to its suitability for generating content for the Cén Scéal? application. The model was found to produce pedagogically unsound material, but it may still be useful in assisting the manual curation of high-quality language learning resources.