The deposit provides a controlled corpus of twenty short narratives co-produced with a large language model under two continuity regimes: Memory ON (continuity enabled) and Memory OFF (continuity disabled). The repository includes the original DOCX narratives (why_a_cat_ON01 – why_a_cat_ON10; why_a_cat_OFF01 – why_a_cat_OFF10), a deterministic Python analysis pipeline, and the derived datasets used to produce Tables 1–6 in the manuscript. Each narrative was generated in a fresh chat session, with the study prompt as the first and only task-specific message, ensuring no local conversational context. Re-running the provided pipeline reproduces all reported numerical results, given the same inputs and package versions. See README for file nomenclature, execution instructions, and outputs.