This paper presents the design, implementation, and evaluation of a conversational system for analyzing and explaining feline behavior using a Retrieval-Augmented Generation (RAG) pipeline. The system integrates two large language models (LLMs) (ChatGPT and DeepSeek) with external knowledge sources indexed in a vector database (Pinecone), allowing user queries to be enriched with relevant contextual information from veterinary articles, PDF documents, and expert content. Through a WhatsApp interface powered by Twilio and Flask, users interact with the system to receive tailored guidance on domestic and feral cat behavior. A total of four system configurations (each LLM with and without RAG) were evaluated using a survey administered to 40 cat owners, who assessed the quality of 400 generated responses. Results indicate that while non-RAG responses were generally preferred, RAG-enhanced replies provided more precise and contextually grounded recommendations in specific cases. The findings underscore the importance of corpus quality in RAG-based systems and offer practical insights for deploying LLMs in animal behavior support tools. This work contributes to the fields of computational ethology, veterinary informatics, and human–animal interaction by demonstrating a scalable, domain-adapted AI solution.