What is AI visibility?

Start with the prompt and answer. Record brand mentions, recommendation order, descriptions, cited URLs, provider, account or location context, and run time. Keep the raw answer so a reader can audit the metric. Define what counts as a mention before scoring, including aliases and names shared by several brands.

Separate answer mentions from citations. A brand can appear without its own site being cited; a source can be cited without an explicit brand recommendation.

Include category, problem, comparison, and purchase-intent prompts that real buyers might ask. Write the list before running the study. Keep prompts stable across repeated runs and record any change.

Group results by topic and product. Report attempted runs, valid captures, and failures beside each rate. Apply the same rule across dates; a changing denominator can distort a trend.

Generated answers vary with wording, account state, location, retrieval, and time. Show the sample size and collection method. Repeat the same prompts and inspect the variation before treating a small change as a trend. These samples do not measure every private user conversation.