Development of an Exploratory Taxonomy for Veterinary Professionals' AI Query Patterns Across Clinical Stages: An Expert Panel Study

Abstract

Background/ObjectivesThe integration of large language model (LLM)-based AI tools into veterinary clinical practice is rapidly increasing; however, no systematically derived taxonomy of veterinary AI query patterns has been established. This study aimed to develop and refine an exploratory taxonomy of veterinary AI query patterns across clinical stages through a structured expert-panel review process. MethodsAn exploratory cross-sectional expert panel study was conducted. 5,372 real-world query logs from a veterinary clinical AI chatbot deployed over eight months were analyzed using AI-assisted inductive coding to derive an initial taxonomy. The taxonomy was refined through literature review and subsequently reviewed by an expert panel of 38 veterinary professionals via structured online survey across five clinical stages. ResultsA taxonomy of 3 categories and 21 subtypes was established: Clinical Support Queries (Types A-H), Evidence-Based Research Queries (Types I-L), and Terminology and Drug Reference Queries (Types M-U). Type B (Differential Reasoning) had the highest overall frequency (57/188 first-choice responses), while Type D (Clinical Decision Support) was dominant immediately post-consultation (55.3%). Veterinary professionals with [≥]10 years of experience showed a higher frequency of Type G (Evidence Search) preference than those with <10 years of experience (18 vs. 4), while university-affiliated professionals demonstrated a distinct pattern dominated by Type G. ConclusionsTo our knowledge, no published study has previously established a veterinary-specific, clinical-stage-sensitive exploratory taxonomy of AI query patterns; this study addresses that gap. The findings provide a foundational framework for designing context-aware, stage-adaptive veterinary AI systems and benchmark evaluation tools.