This study used a written interview format to ask several large language models (LLMs, specifically, Claude, Chat GPT, and Gemini) about their ability to write reading passages restricted to particular vocabulary lists, and with specific restrictions on text complexity. The presentation outlines the background of the study, particularly features that might be included in passage specifications, and provides options for setting specifications in a particular context and/or at a particular level of English proficiency. It outlines the various capabilities claimed by current versions of the LLMs that were interviewed.
