Background: interRAI assessments are mandatory for publicly funded home or residential aged care in New Zealand. These comprehensive geriatric assessments, cover over 300 items to evaluate physical and psychosocial well-being and guide care planning. However, the vast amount of data can overwhelm patients, families, and infrequent users. This research aimed to use Large Language Model prompting to generate and refine patient summaries for use in clinical practice. Methods: A closed Azure environment with 2400 de-identified interRAI assessments (25% Māori) and interRAI item definitions and score interpretation data was used to develop test summaries. Draft summaries were presented to focus groups (health professionals involved in interRAI assessments, older people, families) to review and provided feedback (structured around modified PDQI-9 criteria – a tool originally developed to assess quality of clinical notes). Results: Three rounds of focus groups and refinements were required to produce summaries and care plans that had appropriate language, content and formatting. Further results regarding the types of changes required and variation between health professional and older person/family feedback will be available at the conference. Conclusions/implications: This presentation will focus on the methods used to develop interRAI summaries, examples of AI produced summaries, and the protocol for testing the summaries in a clinical trial to investigate feasibility and acceptability. The learnings are applicable and relevant to health information development and communication more broadly.