Prompt sampling, explained
AI assistants do not have a public results page or a rank tracker. Their answers vary from one run to the next and differ between users, products and models. Prompt sampling is the method that makes AI visibility measurable despite that: choose a fixed set of realistic questions, run them across the assistants that matter on a regular schedule, and record the results the same way each time.
The questions are the foundation. They should reflect how clients actually ask, including specific, conversational and comparative questions, not just short keywords. A good set mixes category questions ("who are the best estate planning attorneys for a family business?"), brand questions ("what is [firm] known for?") and comparison questions. Each result is recorded for presence, position, accuracy of description, competitors named and sources cited.
Because individual answers are noisy, conclusions come from patterns across the set and across time. Running each question more than once, keeping conditions consistent and noting model or product changes all improve reliability. The output is a baseline for AI share of voice and a list of specific gaps to address. Prompt sampling is the core of an AI visibility audit.