Why it matters
A hydraulics distributor runs a fixed prompt set every month and checks whether ChatGPT and Perplexity cite its Parker-to-Gates interchange chart when a buyer asks for the "Gates equivalent of Parker 387 hose." If the chart gets cited, the distributor logs the engine and the date. Do that across 30 cross-reference prompts and you get a citation rate: the share of answers that quote you as the source. That number tells you something keyword rankings cannot. It shows whether the engines treat your catalog as the authority buyers act on, or send them to a competitor's interchange data instead.
Citation tracking vs share of voice
Share of voice counts how often a brand is mentioned in AI answers across a prompt set. Citation tracking is narrower. It counts the subset where the engine names your domain as the cited source, with a link or attribution, not a passing mention. You can be mentioned without being cited. Tracking both tells you the gap between being known and being quoted.
In practice
Build a stable list of the prompts your buyers actually ask: spec lookups, cross-reference questions, "best supplier for X." Run each prompt across the engines several times, since answers vary per run, and record which URLs get cited. An MRO supplier might find its PIM-fed product pages get cited for part-number queries but never for application questions, then route content work toward the gap. Several tools automate the sampling, but the method is vendor-neutral.