This paper presents a prototype system for discovering patterns of household movements using the Google Generative AI (Gemini-1.5) large language model (LLM) and implemented through the Flask web framework. The system processes unstandardized name and address data to identify instances where two or more individuals share addresses at different points in time and subsequently move together to another location. The key advantage of using LLMs is their ability to understand data regardless of syntax quality, variations in address formats, and differences in how names or personal information are recorded. This research shows that an LLM is capable of detecting these patterns and provides a roadmap for automating the extraction of complex relationships between household members. This will improve entity resolution (ER) for address-based movements, and explore the efficacy of generative models in handling unstructured, low-quality data.